To identify the most important determinants of the optimal timing for the initiation of antenatal fetal testing in a high-risk pregnancy (such as insulin-dependent diabetes mellitus), we performed a decision analysis based on a Markov model. The model incorporated several components: test sensitivity and specificity and week-specific probability estimates for fetal death, delivery, and neonatal morbidity and mortality. The analysis demonstrated that the optimal time in gestation to begin testing (from the combined standpoint of both maximizing neonatal survival and minimizing fetal death) was highly dependent on test specificity, the cumulative risk of fetal death (and its distribution throughout gestation), and the week-specific probabilities of neonatal death. For example, even with the increased risk of fetal death associated with insulin-dependent diabetes, very high test specificity (>99%) is required to recommend testing before 30 weeks. If tests of lower specificity (higher false-positive rates) are used, neonatal deaths will increase and will not be offset by a corresponding decrease in fetal deaths. By performing sensitivity analyses, we demonstrated the effect and relative importance of the model components for determining the optimal gestational age to initiate fetal testing. © 1995, All rights reserved.