Methods: The first was an Analytic Hierarchy Process (AHP) decision model, the second was a decision tree, and the third was a hybrid AHP-decision-tree model. Decision criteria for all models included four maternal outcomes - hysterectomy, numbness/pain near previous incision, incontinence and placental abnormalities - and two neonatal outcomes: death and disability. Utilities for the decision tree were derived from a previous study that measured the childbirth preferences using the AHP. Outcome probabilities used in the models were derived from a systematic review and from clinical experts. The major difference between the models was that the AHP model used subjects' perceptions of risk, while the decision tree and hybrid models used the probabilities of risks directly. The AHP and hybrid models also included criteria related to having a desired delivery experience.

Results: The AHP model generated an elective repeat cesarean (RC) decision for 73% of the women, while the decision tree model and hybrid model generated significantly lower rates of cesarean decisions compared with the AHP model, 14% and 18% respectively (p<0.001). The decisions generated by decision tree and hybrid models did not differ significantly (p=0.63). The AHP model used a subjective risk assessment while the other two objectively incorporated absolute risks. For example, subjects overwhelmingly preferred cesarean due to a lower probability of harm to infant as compared to trial of labor (TOL). However, TOL is better from the mother's health conditions' point of view. Since the probabilities of maternal risks are of orders of magnitude larger than the infants' risk probabilities, the decision tree model favored TOL while the AHP model that included patients' subjective responses to risk favored RC.

Conclusions: Decisions about childbirth are challenging to model and discuss as there are health and non-health factors and two patients to consider. We found that the decision tree/hybrid models followed the objective absolute health risks of options whereas the AHP model gave more weight to the perceptions of the mother. Both perspectives are important. The results of these models will be particularly helpful to patient-provider dialogue as the comparison reveals important gaps between patients' perceptions of risk and actual risks.