PubMed1995
For hypertension management, it is increasingly recommended to individualize treatment according to the individual patient's characteristics. However, the ability of physicians to predict the patient's risk in the absence of a computer aid is unknown. Our aim was to determine if senior hypertension specialists were able to assess reproducibly cardiovascular disease (cvd) risk in hypertensive patients. The records of 100 patients were selected at random from the ARTEMIS database. The group comprised 53% of males, mean age was 43.7 (15), blood pressure 160 (25) / 99 (12) mmhg and serum cholesterol 5.6 (1.1) mmol/l. The records were summarized on a standardized form. Nineteen items were devoted to general characteristics and medical history of patients, 15 to clinical examination and 7 to biological tests. These items have been chosen because they were used in the framingham cvd risk prediction equation and/or had decisional value in a bayesian program designed for secondary hypertension diagnosis. Six physicians, twice at one month intervals, assigned each patient to one of the three categories of risk (low, moderate, high). The order of cases was randomized. Chance corrected agreement (intra-physician agreement) between the two estimations was only moderate except for one expert (kappa values: 0.658, 0,543, 0.439, 0.363, 0.328 and 0.207 for each expert, respectively). The proportion of patients assigned by each physician to each category ranged from 22.7% to 77.3% for the low risk, and from 4.5% to 43.9% for the high risk category. One physician overestimated risk, one underestimated it and the three others were in keeping with framingham predictions. Bias and lack of reproducibility influence physician judgment in assessing cvd risk of hypertensive patients. Application of computer aids could result in more rational assessments and better decision making.