Medical Decision Making

This research examined the relative importance of information gathering versus information utilization in accounting for errors in diagnostic decision making. Two experiments compared physicians' performances under two conditions: one in which they gathered a limited amount of diagnostic information and then integrated it before making a decision, and the other in which they were given all the diagnostic information and needed only to integrate it. The physicians. 1) frequently failed to select normatively optimal information in both experimental conditions; 2) were more confident about the correctness of their information selection when their task was limited to information integration than when it also included information gath ering; and 3) made diagnoses in substantial agreement with those indicated by applying normative procedures to the same data Physicians appear to have difficulties recognizing the diagnosticity of information, which often results in decisions that are pseudodiagnostic or based on diagnostically worthless information. Key words. diagnostic reasoning; infor mation gathering; information integration; Bayes' theorem; medical decision making (Med Decis Making 1991;11:233-239)

Standard recommendations for patients who have had superficial bladder cancer are in spection by cystoscopy quarterly for a year or two after tumor removal, then half-yearly and yearly. The authors assessed the potential for improvement in scheduling cystoscopies according to probabilistic optimization techniques. Eight hypothetical practices were created, based on retrospective analysis of 918 bladder-cancer-patient charts. Standard and alter native recommendations for the interval to next cystoscopy were compared. The alternatives were derived from patient-specific predictions of future tumor risks (based on the patient's prior recurrence rate and tumor stage and grade) and a nonlinear optimization approach to allocation of the same number of cystoscopies as were available for standard follow-up. The optimization proposed longer intervals between visits for low-risk patients and shorter inter vals for high-risk patients. Overall, optimization reduced expected tumor detection delays by 30%, from 12.6 to 8.7 weeks. When optimization intervals were shorter than standard, cancer was found more often at subsequent cystoscopies (34% vs 27%, p < 0.05), suggesting that the optimization was a better predictor of cancer recurrence. If reduction in tumor-detection delay is the goal of follow-up for recurrent cancers, then urologists can improve monitoring by using probabilistic optimization methods for scheduling cystoscopies. Further understand ing of the accuracy of predictive models for bladder-cancer recurrence rates is desirable. Subsequently, the optimization method developed here may be tested prospectively. Key words: bladder cancer; nonlinear optimization; operations research; follow-up studies; mon itoring; cystoscopy; health services; decision analysis. (Med Decis Making 1991;11:240- 248)

To derive meaningful conclusions in a changing medical setting, medical decision-support systems must represent and reason about the temporal nature of the clinical environments they attempt to model. Because all difficult medical problems have significant temporal features, designers of medical decision support systems must recognize the unique problems caused by representing and reasoning with temporal concepts. This report has three goals. 1) to describe a set of fundamental issues in creating and reasoning with computer models of a changing clinical environment, 2) to present a taxonomy for characterizing the temporal characteristics of computer models of temporal reasoning, and 3) to use this taxonomy to compare the models of time used in some implemented medical decision-support programs. From this examination, it is argued that computational models of time based on a single uniform representational or inferential method are limited by the expressive power of that method. Multiple modeling formalisms that express different temporal properties of the do main task and that work cooperatively are required to capture the subtlety and diversity of temporal features used in expert clinical problem solving. As an example of this approach, the author describes a program called TOPAZ that contains two temporal models that represent different temporal features of the clinical domain. Key words: temporal reasoning and rep resentation; computer-based problem solving; physiologic models and simulation. (Med Decis Making 1991;11:249-264)

Computer-based clinical simulations for medical education vary widely in structure and for mat, yet few studies have examined which formats are optimal for particular educational settings. This study is a randomized comparison of the same simulated case in three formats: a "pedagogic" format offering explicit educational support, a "high-fidelity" format attempting to model clinical reasoning in the real world, and a "problem-solving" format that requires students to express specific diagnostic hypotheses Data were collected from rising third- year medical students using a posttest, attitudinal questionnaire, students' writeups of the case, and log files of students' progress through the simulation. Student performances on all measures differed significantly by format. In general, students using the pedagogic format were more proficient but less efficient. They acquired more information but were able to do proportionately less with it. The results suggest that the format of computer-based simulations is an important educational variable. Key words. medical education, undergraduate; clinical reasoning; computer-assisted instruction. (Med Decis Making 1991;11:265-272)

Informed consent litigation provides a forum in which probabilistic evidence is elicited from physicians as parties or as expert witnesses. The authors reviewed over 450 medical in formed consent opinions reported by both trial and appellate courts in all 50 states over 40 years to determine 1) the extent to which verbal expressions of probability were used by testifying physicians to characterize the risks of medical procedures; 2) when such expres sions were used, whether consistent numeric interpretations of the terms were being applied by the physicians; 3) whether the choice of expression was influenced by the severity of the consequences associated with the particular risk, and 4) whether the use of such terms was correlated with trial outcomes, inasmuch as the duty to disclose a risk is said to increase with the magnitude of the risk and probability is one measure of such magnitude. It was found that subjective verbal expressions of probability are used in the litigation setting, and that such expressions represent broad ranges of numeric probabilities. There was some correlation between the expression and the represented numeric probabilities. In general, expressions such as "extremely low" and "low" corresponded to probabilities lower than those represented by terms such as "high" and "very high." Further, verbal expressions appeared to be influenced by the seventy of the consequences associated with the risks, but whether this increases or decreases the ambiguity of verbal expressions in the com munication process warrants further research. The authors suggest a syntax of verbal expres sions of probability as a means to reduce the numeric ambiguity of these terms Key words: verbal expressions; probability; syntax; communication; medicolegal, litigation; informed con sent. (Med Decis Making 1991;11:273-281)

A cost-utility analysis was performed to evaluate the relative importance of each of the various elements in the current circumcision debate. Elements used in the analysis included the cost of the procedure, the pain associated with the procedure, the risk of urinary tract infections, and the risk of penile cancer. The net, discounted lifetime dollar cost of routine circumcision is $102 per person, while the net, discounted lifetime health cost is 14 hours of healthy life. These results suggest that the financial and medical advantages and dis advantages of routine neonatal circumcision cancel each other and that factors other than cost or health outcomes must be used in decision making. Key words: cost-utility analysis; circumcision; urinary tract infection. (Med Decis Making 1991;11:282-293)

The authors argue that for risky medical treatment decisions, conventional utility assessment techniques are inadequate due to their emphasis on unrealistic risk magnitudes and sure consequences, leading to assessment questions that are unfamiliar to most patients, have little educational value, and do not reliably extend to the application at hand. As an alternative, medical utility assessments should be performed in a continuous-risk domain with risk levels matching those of the actual decision problem. In support of this position, the authors describe an empirical study that compares the responses of subjects under a conventional assessment protocol with those of subjects under a continuous-risk utility assessment protocol. Prelim inary results show that conventional assessment protocols result in significantly lower esti mates of the degrees of risk aversion compared with a more realistic continuous-risk protocol. Key words: continuous-risk utility assessment; decision making. (Med Decis Making 1991 ;11:294-304)