Harold C. Sox

Clinical prediction rules: applications and methodological standards

The objective of clinical prediction rules is to reduce the uncertainty inherent in medical practice by defining how to use clinical findings to make predictions. Clinical prediction rules are derived from systematic clinical observations. They can help physicians identify patients who require diagnostic tests, treatment, or hospitalization. Before adopting a prediction rule, clinicians must evaluate its applicability to their patients. We describe methodological standards that can be used to decide whether a prediction rule is suitable for adoption in a clinicians practice. We applied these standards to 33 reports of prediction rules; 42 per cent of the reports contained an adequate description of the prediction rules, the patients, and the clinical setting. The misclassification rate of the rule was measured in only 34 per cent of reports, and the effects of the rule on patient care were described in only 6 per cent of reports. If the objectives of clinical prediction rules are to be fully achieved, authors and readers need to pay close attention to basic principles of study design.

On the Elicitation of Preferences for Alternative Therapies

We investigated how variations in the way information is presented to patients influence their choices between alternative therapies. Data were presented summarizing the results of surgery and radiation therapy for lung cancer to 238 ambulatory patients with different chronic medical conditions and to 491 graduate students and 424 physicians. We asked the subjects to imagine that they had lung cancer and to choose between the two therapies on the basis of both cumulative probabilities and life-expectancy data. Different groups of respondents received input data that differed only in whether or not the treatments were identified and whether the outcomes were framed in terms of the probability of living or the probability of dying. In all three populations, the attractiveness of surgery, relative to radiation therapy, was substantially greater when the treatments were identified rather than unidentified, when the information consisted of life expectancy rather than cumulative probability, and when the problem was framed in terms of the probability of living rather than in terms of the probability of dying. We suggest that an awareness of these effects among physicians and patients could help reduce bias and improve the quality of medical decision making.

The potential size of the hospitalist workforce in the United States

PURPOSEnIn the United States, there are currently 1,000 to 2,000 physicians who specialize in inpatient hospital care. The number of such hospitatists appears to be growing rapidly, but the ultimate size of the hospitalist workforce is not known.nnnMETHODSnWe obtained workload data from 365 practicing hospitalists who completed a survey by the National Association of Inpatient Physicians. We then estimated the number of potential hospitalists, based on published national hospital census data. We assumed that hospitalists would care for all medical inpatients, but only at hospitals large enough to require > or = 3 hospitalists. We also made estimates based on the primary care physician referral base and international benchmarks. We estimated hospitalists primary care referral base from telephone interviews with key informants. Official sources in England and Germany provided international workforce data.nnnRESULTSnHospitalists reported an average workload of 13 inpatients. To cover all adult medical inpatients in the United States, we estimate a potential workforce of 19,000 hospitalists. Sensitivity analysis yielded 10,000 to 30,000 hospitalists. Our alternative models yielded estimates within this same range.nnnCONCLUSIONSnThe future hospitalist workforce is potentially quite large. This finding highlights the need to evaluate the economic and clinical outcomes of hospitalist systems.

BREAST CANCER SCREENING

Randomized controlled trials involving nearly 500,000 women on two continents have confirmed the early promise that screening mammography can reduce breast cancer mortality. The observed benefits of mammographic screening, however, are not the same in all women. The mortality reduction in women over age 70 is unknown, and women aged 40 to 49 do not appear to benefit from mammographic screening to the same extent as those over age 50. The reasons for this disparity are incompletely understood, but it depends in part upon differing tumor biology and mammographic test characteristics in younger women. Even if relative survival benefits were equal for women under and over age 50, absolute reduction in risk would remain considerably lower for younger women, a disparity that would not be corrected by improved screening technology or adjustment of interscreening intervals. The authors review of the evidence leads them to strongly support mammographic screening of women aged 50 to 69 at an interval not longer than 2 years. The authors also feel it is reasonable to screen women over age 70 who have a favorable life expectancy. They conclude, however, that the evidence does not support a blanket recommendation in favor of screening women aged 40 to 49. Instead, they advocate a well-informed conversation between physician and patient regarding the present knowledge and the risks and benefits of screening for each individual woman. Definitive answers await the results of ongoing RCTs designed to study the survival benefit conferred by screening women aged 40 to 49. Disagreement will undoubtedly persist regarding which recommendations should determine private practice and public policy.