Beliefs and interests are the lenses through which empirical evidence about risks are interpreted. These beliefs and interests may fundamentally differ amongst, and between professionals. Van Teijlingen (2005) and Davis Floyd (2001) distinguish two approaches to risk in MNC: the social model and medical model of care.[56,57] In contrast to our findings, their conceptualisation does not include the role of interests. Furthermore, although they donot explicitly assign the social model tomidwives and themedical model toobstetricians, they make an implicit division along the lines of professional boundaries. This frame may impair efforts to improve risk selection. For example, it may lead to professional abdication, an act of not daring to speak out against the norm, especially in cases where power is unequally distributed.[58] Under the dominance of the biomedical model of care, this may prevent professionals to openly criticize practices, such as unnecessary interventions or disrespectful care, reinforcing the biomedical model of care. It also may prevent midwives and obstetricians from critical self-reflection on how their beliefs and interests influence their decision making. In our historical study, we found an association between the beliefs and interests of professionals and their education and socialisation. This may partly explain the regional variation found in the studies of the organization of risk selection and in protocols for neonatal referral. However, even if professionals have the same level of risk perception, they may make different choices in individual cases.[59] The variation in assessment of risk generated by personal beliefs and interests may induce bias in risk selection. Individuals’ understanding of risk is actively formed, by looking for, and using information. But they may also make conscious decisions to avoid certain forms of information.[60–66] Bias may be caused, for example, by the tendency to look for confirming evidence to support a hypothesis rather than looking for disconfirming evidence to refute it (confirmation bias), the tendency for things to be judged more frequently if they come readily to mind (availability bias), and matching signs and symptoms to what is known (representativeness bias).[67–69] These biases lead to professional overconfidence. [67,68,70,71] Risk researchers argue that bias is inherent to human considerations of risk.[67,68,72–76] Bias in risk selection is confirmed by numerous studies, showing that perception of risk is not neutral, but complex andmultifaceted,[26,32,53,77–84] influenced by many factors, including personal, professional, social, political, cultural, organizational, ethical and educational factors.[67,77,85–97] Risk selection is prone to bias as there is always some degree of uncertainty.[68,71] From a risk theory perspective, certainty about risk is not possible.[22,98] All risks are uncertain. [10,28,99–101] Risks might be unknown, known with insufficient precision,[102] or they may not clearly be high or low.[103–105] Even if they can be calculated, they will remain uncertain due to the heterogeneity in study populations and unavailable knowledge. 165 7 GENERAL DISCUSSION