II. Surveying the Evidence of Bias in Medical Decisions

A. Defining the Inquiry

This Part of the Article examines the evidence, primarily found in the medical literature, that suggests that a physician's choice of medical diagnosis and treatment for an individual patient may, in some cases, be influenced by patient characteristics that are unrelated to the patient's need for medical care. This evidence appears most often in studies identifying disparities in the rates of a particular intervention received by patients according to group-related characteristics such as race or gender. However, any inquiry into the meaning of group-based disparities in the rates of health services must be undertaken with awareness of the recognized pervasiveness of variations in medical practice.

Since the early 1970s health services researchers have demonstrated that the rates of various procedures vary widely by geographic location. These varying practice patterns do not reflect different rates of illness or medical need by geographic region; instead, researchers have attributed them primarily to differences in physician practice styles. In other words, the explanation is not that patients are sick differently in different regions, but that doctors in different regions treat similarly sick patients differently. On reflection, these geographic variations are perhaps not surprising; variations in the rates of services are predictable as long as medical practice is an inexact science. When the appropriate intervention for a patient is not clear, the physician exercises some discretion in making a treatment choice, and either the traits, training, background of the *202 individual physician or characteristics of the medical community can influence those choices.

Before beginning to examine the possible role of bias in medical treatment decision making, it is critical to isolate the phenomenon targeted for examination. Geographic variations research focuses on the role that geography plays in influencing treatments provided. Other physician-related characteristics, such as physician age, specialty, training and gender may also influence medical decision making and practice styles. For purposes of the present inquiry, the role of these physician-related characteristics must be distinguished from the role of patient-related characteristics in medical decision making. Differentiating between the role of physician-related characteristics and that of patient-related characteristics can be complicated. In some cases, a characteristic of the physician might correlate with a higher likelihood that the physician's medical judgment will be biased by a characteristic of the patient.

Another distinct factor that is largely extrinsic to either the physician's or the patient's makeup may also influence a physician's treatment choices. As public and private health care financing programs have increasingly incorporated managed care methodologies, many physicians practice under reimbursement systems that reward physicians financially for ordering fewer and less expensive interventions. These financial incentives are designed to influence physician practice styles and a robust literature examines the predictable conflict of interest that arises when a patient's best medical interests are contrary to the physician's financial interests.

A final characteristic that may influence a physician's decision whether and how to treat a patient is the patient's own financial ability or inability; this patient attribute, however, lies beyond this Article's focus. Although financial status is a patient characteristic that is not directly related to the patient's need for medical treatment, a decision to refuse to treat a patient who cannot pay for medical services or to limit a patient's treatment options to those for which he has the means to pay is generally *203 (though by no means universally) deemed legitimate by both the legal and medical communities. Although codes of medical ethics call upon doctors to provide charity care, they do not require a particular doctor to treat every indigent patient who seeks care; nor does the legal system generally impose on doctors the obligation to provide uncompensated care. Differential rates of health insurance coverage undoubtedly play some role in contributing to disparities in levels of care received by different population subgroups. Indeed, some evidence suggests that physicians may harbor a bias against treating “the poor” independent of particular poor patients' health insurance coverage. Nonetheless, while acknowledging the argument that poverty may represent a group-related characteristic that can engender stereotypical attitudes, the difficulty of distinguishing between the acceptable and the unacceptable influences that medical indigency may exert on medical treatment decisions leads me to exclude financial ability and insurance status from the sources of bias on which this Article focuses.

So what is left for examination? This Article's inquiry will focus on the role that patient characteristics, not related to the patient's need for medical treatment, play in influencing a physician's choice of diagnostic or therapeutic intervention. I phrase the question in general terms to emphasize that this is not an article about racist medical decision making per se (or sexist, or ageist, or ablist, or heterosexist or whatever “-ist” medical decision making). Instead, it is a broader examination of how the legal system does or could respond to cases in which a physician's medical judgment is somehow infected or biased by a non-medical characteristic of the patient.

But even with so cabined an inquiry, assessing when a patient characteristic is indeed unrelated to the patient's need for medical care may sometimes prove tricky. Group-related characteristics such as race, ethnicity or gender may not generally be related to a patient's medical needs; indeed, it has been argued that medical research that seeks to attribute particular outcomes to patient race is irrelevant to treatment choices for *204 an individual patient. By contrast, a committee of the Institute of Medicine took the stance that demographic characteristics may sometimes be reliable proxies for factors such as patient preferences or anticipated outcomes of care; in those instances, such characteristics might appear to be related to the patient's medical needs. Similarly, as psychiatrist Sally Satel asserts, “[when] certain diseases and treatment responses cluster by ethnicity, [r]ecognizing these patterns can help [doctors] diagnose disease more effectively and prescribe medications more effectively.” Crucial to the question of relatedness to a patient's medical needs is the reliability of the proxy a physician employs in making medical judgments. For example, if clinical research demonstrates that females of reproductive age consistently metabolize a drug at a faster rate than males, then female sex is related to a judgment regarding the drug's proper dosage for a female patient. By contrast, if a physician simply assumes, without scientific support, that females are better able than males to endure pain, the patient's sex should be seen as an unreliable proxy for--and thus not related to--the level of an individual patient's medical need for pain relief.

The high level of uncertainty that still permeates much of medical practice complicates the “relatedness” question. To illustrate, one study examined whether a patient's race influenced the frequency with which nephrologists diagnosed patients with end-stage renal disease as having hypertensive renal disease, as opposed to some other form of underlying kidney disease. The researchers' findings suggested that doctors may have been more likely to label African Americans' kidney disease as hypertensive than they were to give white patients a similar diagnosis, even when the patients had similar clinical histories. The researchers concluded that some nephrologists in the study appeared to use race as a diagnostic factor even though the diagnostic validity of that factor is not clear. The researchers cautioned that the use of such unvalidated diagnostic criteria *205 could produce a “self-reinforcing truth.” Clinicians employing race as a diagnostic preference may overdiagnose African Americans as having hypertensive renal disease; these diagnoses will be incorporated in epidemiological data regarding disease incidence among racial groups; these data may then be read as supporting the use of race in diagnosing kidney disease because of the high incidence of hypertensive kidney disease among African Americans. Thus, any assumption that a demographic characteristic is related to a patient's need for medical treatment should be carefully scrutinized to ascertain its scientific validity.