- We present two models that guide clinicians and parents who must decide whether to perform amniocentesis (AC) for detection of fetal abnormalities: a personal utility model for parents, and a cost-benefit model for a health-care provider. These models can tailor general guidelines to specific AC decisions. The American College of Obstetricians and Gynecologists (ACOG) provides separate guidelines, mainly concerning the use of AC and of high-resolution ultrasound (HRUS), for detection of Down's syndrome (DS) and neural-tube defects (NTD's). We used influence diagrams to model two decisions: (1) the parent's decision, using preference probabilities as personal utility measures; and (2) the decision of a policy-making organization, such as a large health-care organization, using monetary utility measures for test costs and for the values of life and disability. In both cases, we tailor the model to the particular case by using parent-specific prior probabilities for DS and NTD, as well as the probability for miscarriage due to the AC procedure. The results of the analysis indicate significant sensitivity of the AC decision to parent-specific utilities. Such utilities have been elicited by researchers such as Pauker and colleagues, and have been shown to vary widely. The AC decision also is sensitive to the rate of miscarriage caused by the AC procedure, but is sensitive to the DS prior probability (e.g., by age) only for certain personal-utility values. The NTD's prior probability plays a significant role mainly in cases where the decision is otherwise not obvious. In the case of the model using monetary measures, break-even points for the health-care provider can be shown for all three case-specific prior probabilities, given the value of life and the cost of lifelong disability. The AC decision is affected by parent-specific personal utilities and prior probabilities for DS and NTD's, and by local rates of miscarriage due to AC. Monetary utility measures can be used in the case of a large-scale policy. Both measures of utility might be combined, maximizing societal and individual utilities. Using only arbitrary cutoff values (e.g., age), considering only one disorder at a time, or including only one test procedure oversimplifies the AC decision. Additional nodes, signifying relevant parameters and utilities, can be added with relative ease to both of our influence-diagram models. We demonstrate the need for both comprehensive data and parent-specific utilities by a four-way sensitivity analysis and several exemplary cases.