Search PubMed⌕ Search

Biomedical subjects

D Spiegelhalter

Publications and source records attributed to D Spiegelhalter.

27 records · Page 2Linked to original sources
PubMed1984

To catheterise or not to catheterise? An approach based on decision theory.

To determine whether or not patients require cardiac catheterisation before surgery a computer based mathematical model was constructed based on decision theory. The model was specifically applied to sick infants under 3 months of age with suspected coarctation of the aorta, and a three way sensitivity analysis was carried out to assess the effects on the model of changes in the probabilities that underlie the decision itself. The optimal decision (that with the greater survival rate) was moved away from cardiac catheterisation to confirm the diagnosis towards operating without cardiac catheterisation by the following factors: a higher probability of survival of operation both in the presence and absence of coarctation; a higher probability of survival if there was no coarctation and no operation performed; a lower sensitivity of catheterisation; a greater incremental risk of operation resulting from previous catheterisation; and a higher relative risk of catheterisation in patients without as opposed to with coarctation. Factors that tended significantly to move the decision towards catheterisation to rule out coarctation rather than neither to operate nor to catheterise were: a lower risk of surgery for coarctation if present; a higher risk of failing to operate on a patient who had coarctation; a high specificity of cardiac catheterisation; a lower incrementation of surgical risk by previous cardiac catheterisation; and a lower relative risk of catheterisation if coarctation was absent. In this institution, the model argues strongly against cardiac catheterisation in the great majority of sick infants with coarctation.

Aortic Coarctation↗
PubMed1983

Analgesic intake and the risk of acute upper gastrointestinal bleeding.

A comparison of aspirin and acetaminophen consumption in 346 matched pairs of patients with hematemesis and melena and general population control subjects has shown associations between intake of both drugs and bleeding. The association for acetaminophen was weaker and was not detectable for habitual intake, whereas the association for aspirin was detectable for recent and habitual intake. Taken overall, the results suggest, by reference to acetaminophen, that patients frequently take analgesic drugs because of the symptoms of bleeding, and such intake is not causal of bleeding. Investigators who fail to employ appropriate control subjects to take account of this point exaggerate the possible risks of aspirin consumption.

Acetaminophen↗
PubMed1983

The value of clinical, biochemical, ultrasound and liver biopsy data in assessing patients with liver disease.

To determine the value of clinical, biochemical, ultrasound and liver biopsy data in the management of patients with liver disease, eight doctors each assessed 75 case histories. With clinical and biochemical data alone, the predictive accuracy was significantly higher when identifying patients as 'medical' rather than 'surgical' (97 compared with 79%, p less than 0.001). However, when making a specific diagnosis as opposed to classifying into medical and surgical categories, clinical and biochemical information resulted in a much lower predictive accuracy for both medical (67%) and surgical (56%) patients. With ultrasound data the predictive accuracy increased to 91% when identifying patients as 'surgical'; with liver biopsy it increased to 99% when identifying patients as 'medical'. The value of the different data assessed involves more than an evaluation of diagnostic accuracy, and in this study the relative worth of each test was therefore assessed on a five point scale based on the effect of the information on the doctors. This included a willingness to give specific treatment and make a specific diagnosis, as well as classifying patients into medical and surgical categories and the confidence they felt in their diagnoses. After clinical, biochemical and ultrasound information the doctors were only prepared to give specific treatment to 11.9% of the medical and 9.3% of the surgical patients. After liver biopsy data, however, they were willing to give specific treatment to an additional 66.6% of the medical patients and 25% of the surgical patients. Further evidence of the value of liver biopsy information came from an analysis of the changes in the doctors' confidence in a diagnosis. Thus, 96 patients were assigned a correct specific diagnosis with clinical and biochemical data alone but none were considered as 'definitive' by the doctors; when liver biopsy information was provided 59 (61%) were placed in this category.

Biopsy↗
PubMed1982

Aspirin, paracetamol, and haematemesis and melaena.

Aspirin and paracetamol consumption have been compared in 346 matched pairs of patients with haematemesis and melaena, and control individuals in the general community. Both aspirin and paracetamol intake were more common in patients than in controls, but the association for aspirin was stronger and was apparent with both recent and habitual intake, whereas for paracetamol the association was not detectable for habitual intake. The results for paracetamol suggests that patients with bleeding take analgesic drugs in part because of symptoms associated with bleeding, and such intake is not necessarily causal of bleeding. Failure to control investigations to take account of this point has exaggerated the possible risks of aspirin consumption.

Acetaminophen↗
PubMedDate not supplied

Evaluating medical expert systems: what to test and how?

Many believe that medical expert systems have great potential to improve health care, but few of these systems have been rigorously evaluated, and even fewer are in routine use. We propose the evaluation of medical expert systems in two stages: laboratory and field testing. In the former, the perspectives of both prospective users and experts responsible for implementation are valuable. In the latter, the study must be designed to test, in an unbiased manner, whether the system is used in clinical practice, and if it is used, how it affects the structure, process and outcome of health care encounters. We conclude with proposals for encouraging the objective evaluation of these systems.

Diagnosis, Computer-Assisted↗