General practitioners' decision-making on drug treatment of hypercholesterolaemia

Abstract

Drug treatment of elevated cholesterol values may be indicated for individuals at high risk for cardiovascular disease and as a complement to lifestyle advice. Guidelines recommend a numerical risk calculation as the basis for selecting the individuals at highest risk among those who are free from previous cardiovascular disease (primary prevention). Guidelines and tools for risk estimation are often not available or not used, which leads to a risk for over- and undertreatment.

The purpose was to examine General Practitioners' (GPs) decisions on chug treatment for patients with hypercholesterolaemia. Written patient cases were presented either in a paper format (Studies I and II) or on a computer screen (Studies III-V). In Study 1, 38 GPs rated their inclination for drug treatment for 40 case descriptions. Each doctor's strategy was defined as the set of statistical regression weights for the different variables describing the patient. The strategies varied widely between doctors. The most important variable was previous coronary heart disease, followed by the degree of cholesterol elevation. The majority of GPs used two or three of the eight variables in their judgements. In Study II the ability of GPs and medical students to make risk estimates without using risk assessment tools was examined. Both groups underestimated risk, especially for high-risk patients.

In studies III-V, think-aloud technique was used to examine the thought processes leading to the decision to prescribe or not. The GPs were instructed to say aloud all their thoughts about the information presented on successive computer screens. The audio-taped and transcribed protocols were coded for whether the statements favoured or disfavoured drug prescription (directionality of decision). They were also coded regarding both the information about the patient to which they referred and their cognitive content. In Study III, half of the participants not only talked aloud but also rated their inclination toward drug prescription at each of the successive screens with new information. The ratings and verbal protocols reflected the change in directionality in similar ways, which was interpreted as supporting their validity. Verbal protocols were at least as sensitive as ratings in reflecting the change in decision directionality over time.

In studies IV and V the think-aloud data were analysed regarding how different kinds of information about the cases were evaluated and used in the decisions. Cholesterol level was most important and different cut-off levels were used by different GPs. Lifestyle-related factors seemed to be evaluated from different perspectives. A patient's smoking could be regarded as increasing the risk and thereby favouring drug prescription by some GPs, and as a possibility for lifestyle change and as an argument for refraining from treatment by other GPs. After the six cases, the GPs were asked to describe in their own words their thinking concerning treatment of hypercholesterolaemia. These protocols were coded for cause-effect relations to map the doctors' knowledge and opinions. The GPs made several departures from a strict application of guidelines in the individual cases, even when they had expressed the relevant knowledge contained in the guidelines. Their arguments often concerned lifestyle factors.

The think-aloud technique with the coding scheme developed for the present studies seemed to be a valuable complement to statistical approaches to judgements and decisions (Studies I-II). One possible application area is studies on differences in decision processes of experienced doctors and medical students, the results of which can be used in medical training. Another application area involves the usability and efficiency of different decision support systems integrated in the computerised medical record.