You are on page 1of 2

FIGURE 10*3 Simplified decision tree for (he decision of whether or not to use a n intravenous thrombolytic agent in a patient

with a suspected acute myocardial infarction. The square node represents the decision point, and the round nodes d enote chance events (see text for details). heart failure averted because an infarction was smaller, affect the quality of f uture years of life.; The results and usefulness of a decision analysis depend on the probabilities an d utilities'that are used in the calculation, and it is imperative for decision analyses to include a Sensitivity analysis, in* which various estimates for each probability are included'in the analysis to determine if the conclusions, would be changed. For example, in the analysis in Fig. 10-3, some range of probabilit ies must be assumed for the risk of serious bleeding complications and for the l ikelihood that any particular patient is having an acute myocardial infarction. Because of the impressive benefit ofihromboly- sis for improving survival in pat ients with acute myocardial infarctions, decision analyses have shown treatment to be the preferred option even when an acute myocardial infarction is not certa in and even in most elderly patients. If the conclusions of an analysis were alt ered by relatively minor changes in the assumptions on which it was baseff, the analysis would not be sufficiently reliable to become the basis for decision mak ing. Decision analysis sometimes demonstrates a clear and dramatic advantage with one particular option. In other circumstances, there may be little difference betwe en two options; either option may be reasonable, or secondary issues that cannot be taken into account in the formal analysis, such as the patient's feelings ab out taking risks or the rcccnt local experience with particular interventions, s hould be the final determinants in the decision. Physicians who perform a decisi on,analysis therefore must determine the probabilities of each of the possible e vents by reviewing the pertinent patient experience at their own institution or practice or by reviewing the pertinent literature. Even when the outcome of the analysis seems clear, the physician or the patient may believe that the situatio n in question is <im exception i<i (lie rule, /'urfhcniiorc, even Ihc hcsf aiulyio, tike all clin ical intuition, arc based on assumptions that may be open to debate. In the preceding example, dealing with whether or not to use a thrombolytic agen t in a patient with a suspected acute myocardial infarction, decision analysis i ndicated (he preferred strategy in fcntts of outcome but did not consider the co sts at which such benefits mi^lil be* achieved, If\ determining health policy, a formal cost' effectiveness analysis can be performed to determine how many dolla rs mus( be spent to achieve a unit of benefit, often defined as 2 life saved, a year of life saved, or a quality-adjusted year of life+ saved, in which the year s arc adjusted to take into account the quality of life during that time. For ex ample, in 1994 dollars. 1 year of in- center hemodialysis can be estimated to co st about $35,000 to $45,000 per quality-adjusted year of life saved; this figure includes only the direct medical costs and not indirect costs related to issues such as time lost or travel or any benefits in terms of a patient's ability to work. In some situations, (he ability of the patient to maiotam jzaintbi employm ent may offset some or all of the direct medical expenses. In other situations, such as with pneumococcal vaccine, the saving* from episode s of pneumonia that are averted may more than offuct the cost of the vaccine in high-risk persons. ETHICS AND PATIENT INPUT, In both quantitative and nomjuan- titative clinical re asoning, the physician must consider ethical issues as well as the patient s value s and preferences. While a discussion of these issues is beyond the scope of thi s chapter, it is important to emphasize, (hat patients preferences lor alternativ e therapies may not agree with the preferences (hat the physicians propose on th e basis of (heir own clinical judgment or the result* of a dccision-analysis app roach. For example, many patients with carcinoma of the larynx may prefer radiat ion therapy, with a lower - cure rate but a higher likelihood of maintaining spe

ech, to extirpative surgery. I( is impcra(ive that physicians assess those chara cteristics of life * which the patient prizes most (the elusive quality of life ) p rior to basing controversial decisions solely on quantitative approaches, the ph ysicians own subjective impressions of the likely incdical benefits, their own pe rsonal preferences, or their assumptions about the patient s preferences. While qu antil^ivc analyses may apply to groups, judgment must be exercised w hen adapting them to the individual patient. Therefore, the final plan should re flect an agreement between a well-informed patient and a sympathetic physician w ho has detailed knowledge of the relevant medical issues and of the impact of th e various possible outcomes on the specific patient. REFERENCES American College of Physicians: Guidelines for counseling postmenopausal women a bout preventive hormone therapy. Ann Intent Med 117:1038, 1993 Gaule CB et ai: P neumococcal vaccine: Efficacy and associated cost savings. JAMA 264:2910, 1990 GARNICK MB: Prostate cancer: Screening, diagnosis, and management. Ann Intern Me d 118:804, 1993 Goldman L et al: A computer protocol to predict myocardial infarction in emergen cy department patients with chest pain. N Engl J Med 318:797, 1988 Ka.smki it*: Diagnostic reasoning. Ann Intern Med 110:893. 1989 Lemiio NJ cl al: Bedside diag nosis of systolic murmurs. N Engl J Med 3IX: 1572, l$S Mandei. JS el al: Reducing mortality from colorectal cancer by screening for fccal occult blood. N Engl J Med 328:IJttf. IW Sackett DL cl al: Clinical Epidemiology: A Basic Science ft* 1 7hwi.\AIoInw. 2d ed. Boston, Little, Brown. 1991 Sloand EM cl al: HIV testing: S tale of the an. JAMA 266:2X61. I9VI Sox HC Jr (ed): Common Diagnostic Tests: Use and Interpretation, 2d cd. Philadelphia. American College of Physicians. 1990