Design, construction and evaluation of systems to predict risk in obstetrics.

We present a systematic, practical approach to developing risk prediction systems, suitable for use with large databases of medical information. An important part of this approach is a novel feature selection algorithm which uses the area under the receiver operating characteristic (ROC) curve to measure the expected discriminative power of different sets of predictor variables. We describe this algorithm and use it to select variables to predict risk of a specific adverse pregnancy outcome: failure to progress in labour. Neural network, logistic regression and hierarchical Bayesian risk prediction models are constructed, all of which achieve close to the limit of performance attainable on this prediction task. We show that better prediction performance requires more discriminative clinical information rather than improved modelling techniques. It is also shown that better diagnostic criteria in clinical records would greatly assist the development of systems to predict risk in pregnancy.

The design, construction, operation, and maintenance of the Replicated Modular Estuarine Mesocosm

Perhaps the most difficult job of the ecotoxicologist is extrapolating data calculated from laboratory experiments with high precision and accuracy into the real world of highly-dynamics aquatic environments. The establishment of baseline laboratory toxicity testing data for individual compounds and ecologically important and field studies serve as a precursor to ecosystem level studies needed for ecological risk assessment. The first stage in the field portion of risk assessment is the determination of actual environmental concentrations of the contaminant being studied and matching those concentrations with laboratory toxicity tests. Risk estimates can be produced via risk quotients that would determine the probability that adverse effects may occur. In this first stage of risk assessment, environmental realism is often not achieved. This is due, in part, to the fact that single-species laboratory toxicity tests, while highly controlled, do not account for the complex interactions (Chemical, physical, and biological) that take place in the natural environment. By controlling as many variables in the laboratory as possible, an experiment can be produced in such a fashion that real effects from a compound can be determined for a particular test organism. This type of approach obviously makes comparison with real world data most difficult. Conversely, field oriented studies fall short in the interpretation of ecological risk assessment because of low statistical power, lack of adequate replicaiton, and the enormous amount of time and money needed to perform such studies. Unlike a controlled laboratory bioassay, many other stressors other than the chemical compound in question affect organisms in the environment. These stressors range from natural occurrences (such as changes in temperature, salinity, and community interactions) to other confounding anthropogenic inputs. Therefore, an improved aquatic toxicity test that will enhance environmental realism and increase the accuracy of future ecotoxicological risk assessments is needed.