Dr. Feinstein's excellent article [1] concisely summarizes the various quantitative approaches currently used to define and evaluate clinical judgment. It is true that quantitative models have been used extensively in basic research, and such models are being increasingly used in clinical research studies. I agree that algorithmic outlines graphically showing "if/then" decision making are useful. It must be noted, however, that they are based on the assumption that linear modeling using such mathematical expressions as "0" and "1" is appropriate for all decision-making processes. This approach is somewhat simplistic.

Given the reality that most older patients have multiple chronic diseases, realistic algorithmic presentations of clinical decision making would need to portray the convoluted reality of "but what if" algorithms. Imagine how a "but what if" algorithm would look for the management of a patient with hypertension and diabetes or diabetes and rheumatoid arthritis, or, more to the point, for the management of a patient with hypertension, diabetes, and rheumatoid arthritis?

Although current medical algorithms show the logic behind the management of a particular disease (which is sufficient for a specialist treating one particular disease or system), "but what if" algorithms would realistically reflect the complexity behind the clinical judgment process needed to comprehensively manage patients with multiple disorders. In essence, the practice of internal medicine requires clinical judgment that encompasses a synthesis of vertical (logical) and lateral (creative) thinking skills. It is probably why clinical judgment has thus far been so difficult to define.

Current scientific thinking, however, is beginning to acknowledge the shortcomings of existing linear mathematical models to explain natural phenomena. New theories include "chaos and complexity" models to explain the existence of nonlinearity [2] in physical, biological, economic, and social systems [3] and the "fuzzy set" theory [4] to explain the linguistic inexactness of human reasoning. Mathematical models being developed to support such theories may eventually be useful in explaining the clinical judgment process described by Dr. Feinstein.

Although mathematical approaches do provide the comforts of a common language with which to communicate scientific findings that address uncertainty at many levels, the evaluation of such clinical judgments as prognosis, diagnosis, and treatment will remain elusive unless routine autopsies are done. Autopsies probably provide the most objective and definitive evidence of whether the prognosis, diagnosis, and treatment of a patient were correctly done.

The current rate of autopsies is poor (15%) and suffers from selection bias [5]. Although autopsy is an emotional topic that people avoid because of its association with the unpleasantness of death, I believe that routine autopsies would do much to reduce scientific uncertainty and to provide the objective (hard data) evidence that medicine needs to support and evaluate the rationales of current practice. Given the growing economically based processes used in outcomes research, routine autopsies may eventually become a factor in reimbursement decisions.