Economic Analyses in Clinical Decision Making

The marginal cost is the cost of providing (or withholding) an additional unit of service. This cost is often one of the most relevant for an individual clinician’s medical decision making and is typically quite different from the overall cost allocated to that service. For example, a hospital may have determined that $50 is the cost of providing a chest radiograph. However, a clinical protocol to better identify patients requiring radiographs that resulted in one fewer chest radiograph a day (with no change in outcome) would not “save” the hospital $50 because personnel and overhead expenses would be unchanged; only the expense of any film would be eliminated. Hence the marginal cost to the hospital of one chest radiograph is essentially the cost of one piece of film (even less if digital capture techniques are used). Note that marginal cost varies with volume in a quantum fashion; adding or withholding a larger number of radiographs would at some point dictate a change in personnel and perhaps radiograph equipment, resulting in a different marginal cost. Additionally, the marginal cost is different for the payors and patients; withholding one chest radiograph would save the payors the entire amount they typically reimburse for that radiograph, a figure far higher than the hospital’s marginal cost. Patients would save the cost of their co-pay, if any.

Change in raw mortality is the most common way to evaluate effect on death. In more complex analysis, death and disability are often evaluated in combination as the quality-adjusted life year (QALY); treatment that results in an additional year of life at 100% of normal functioning is credited with 1 QALY; treatment that results in an additional year of life at only 75% functioning is credited with 0.75 QALY.

QALY is more difficult to apply to discomfort, but some believe it can be estimated by the time tradeoff method: A person estimates how many years of discomfort would be acceptable vs a shorter period of perfect health. For example, if a person would prefer 9 years of health to 10 years of chronic pain (but would prefer the 10 years of pain to only 8 years of life), then each year of life with that particular pain is credited with 9/10 = 0.9 QALY. All such QALY estimates are somewhat problematic because people vary widely in risk tolerance and acceptance for various outcomes.

The number needed to treat (NNT) or harm is another way to quantify patient outcome; NNT is the reciprocal of the absolute change in a dichotomous outcome (death, disability). Thus, if a medication causes a 3% net decrease in mortality, 1/0.03 = 33.3 patients need to be treated to prevent 1 death.

The number needed to harm is similar. Thus, for a medication that causes leukopenia in 8% of patients, 1/0.08 or 12.5 need to be treated to harm 1 person. Phrased another way, 1 person is harmed for every 12.5 treated.

The pertinence of the NNT is clearer when comparing mortality to mild adverse effects of a treatment. It becomes cloudier when comparing the reduction in a particular morbidity to a more serious adverse effect. From the perspective of the clinician, however, it can be a very useful tool in explaining the risk:benefit ratio of a treatment to the patient.

Because the NNT is derived from absolute change, as opposed to relative change, it is more clinically relevant to a given patient. For example, a treatment that reduced mortality from 2% to 1% reduces relative mortality by 50% but absolute mortality by only 1%. This is easy to see when represented as an NNT of 1/0.01 = 100 treated to prevent one death. The relative change concept is more relevant to hypothesis testing (proof of concept that a therapy has efficacy) than to an individual patient.

Even when appropriate outcome measures are used and analyzed correctly, it is critically important to note that a statistically significant outcome in a clinical study (ie, with an excellent p value) is not necessarily clinically significant to a patient. Statistical significance is to a large extent dependent upon the sample size; with a large enough sample, a minimal difference of no clinical importance to a patient (eg, a reduction of duration of upper respiratory infection symptoms from 7 to 6.5 days) could well be statistically significant. The magnitude of difference between 2 groups in a clinical study is referred to as the effect size; as in the above example, an effect size may be small, but still highly statistically significant.

The marginal cost-effectiveness ratio is the additional cost of a strategy divided by the additional health outcome it achieves and thus pertains to the situation in which there is a choice between ≥ 2 effective management strategies. Greater health improvement for a given resource expenditure is derived when the ratio is lower.

For policy analysis, the most common measure of effectiveness is the QALY, making the units of the corresponding marginal cost-effectiveness ratio “additional dollars spent per QALY gained.” However, the marginal cost-effectiveness ratio has been criticized because older patients or patients with life-limiting comorbidities have a smaller potential gain in survival from a treatment and therefore have a higher (less advantageous) cost-effectiveness ratio.

For example (see table Calculating a Marginal Cost-Effectiveness Ratio, Analysis 1), consider no antiarrhythmic therapy vs prophylactic use of an implantable cardioverter-defibrillator (ICD) for patients who have survived several months after an acute anterior MI and who have a mildly depressed ejection fraction (between 0.3 and 0.4). (All figures and costs in this example are hypothetical and only for purposes of illustration.) Both strategies assume similar baseline costs for routine care ($78,300), but the ICD has an additional (marginal) cost of $53,100, based on the cost of the device and professional fees, initial hospitalization, and ongoing therapy (including extra physician visits, laboratory tests, medications, rehospitalizations for ICD-related complications, and replacement of ICD generator or leads). If patients with an ICD have a slightly increased life expectancy (7.87 vs 7.42 QALY), the marginal effectiveness of ICD therapy is 7.87 − 7.42 = 0.45 QALY. Thus, prophylactic ICD enhances survival compared to no antiarrhythmic therapy at a cost of $53,100/0.45 QALY, or $118,000/QALY.

Calculating a Marginal Cost-Effectiveness Ratio