Although by most clinicians, risk is viewed upon as chance, in decisional processes, risk describes the combination of probability and impact of an event. In designing and constructing guidelines, risk matrices could be pivotal to present data objectively, to structure group interactions and to come to balanced choices [16]. Before deciding upon new interventions, we first investigated the professional’s opinion on risk reduction by newly introduced interventions and thresholds. The participants opinioned that newly introduced interventions and thresholds should decrease the risk of an event with two to three risk categories. Furthermore, non-invasive interventions should reduce risks on population level, not only in the highest risk group. Nonetheless, despite these expectations on risk reduction, calcium supplementation and a higher cut-off value for diagnosing SGA were introduced. This indicates that decisions made to delineate guidelines are affected by changes in multi-dimensional risk, but apparently also depend on several other factors, in which lowering the most extreme unwanted but preventable outcome, financial aspects, practical consequences for daily practice were most dominant.
Besides the actual risk, risk perception also influences decision-making. Risk perception refers to the people’s subjective assessment of the probability of a specified type of accident happening and its possible impact [17,18,19]. The subjective judgement of risk includes personal beliefs, attitudes, experiences and feelings, irrespective of their validity, but also involves people’s social and cultural background [17, 19, 20]. Perception of risk does not discriminate between risk knowledge on the one hand and the value judgement about its acceptability or tolerability on the other hand [20]. People usually underestimate continuous everyday risks as they are less salient compared with dread risks that may be overestimated, even if both cause the same number of fatalities [17]. Several factors can bias risk perception, such as memorability, age, or media coverage of an event [17]. Overall, important in assessing risks is the ease with which people can imagine how risk materializes in daily practice [21]. Experience with specific risks and memories in specific situations contributes to a more realistic risk perception [17].
We presented both interventions in two separate risk matrices, to show the effect of each intervention to experts that are constructing professional guidelines. The risk matrix without the intervention is seen as reference point. The effect of calcium on the prevention of preeclampsia is less evident than the effect of changing the threshold for SGA. Nevertheless, it took less time for the professionals to decide upon calcium supplementation to be implemented into the guidelines, rather than reaching consensus on the threshold for diagnosing SGA. During the decision process, it became clear that for each intervention different aspects were relevant. On the one hand, despite the absent change in population risk category with and without calcium supplementation, extra calcium given to all pregnant women is expected to cause a substantial reduction on population level in the incidence of preeclampsia as well as related health care costs [8]. Moreover, calcium supplementation is viewed upon as relatively inexpensive intervention, side-effects are uncommon and if any usually mild, and the professionals in our project classified calcium as a non-invasive intervention [8]. On the other hand, by changing the threshold for SGA from the 10th to the 5th centile, the population risk lowered, but at the expense of more offspring complications. Despite a lower population risk, the professionals decided to set the threshold for fetal growth restriction at the 10th centile. This indicates that other factors, in addition to the populations risk level, were important in this decision-making process. On the one hand, decreasing the diagnostic threshold results in more severe neonatal consequences that could have been prevented. On the other and, setting the threshold at the 10th centile results in intensified surveillance shifting care from midwife-led to gynecologist-led. In deciding what to recommend in the guidelines, numerous factors different from population risk and its extremes were mentioned. First, although an ultrasound was viewed upon as minimal invasive procedure, the risk for parent’s uncertainty that comes with the suspicion SGA and the subsequent medicalization was seen as an undesired side effect. Second, as the professionals decided to diagnose SGA definitively after two subsequent ultrasounds that are below the chosen threshold with a 2 week interval, professionals perceived this period of uncertainty quiet long for the future parents [14]. Third, although not easily mentioned by professionals, in the Dutch reimbursement system, referring more pregnant women from midwife-led to obstetrician-led care would have negative impact on midwifes income and consequently a positive impact on hospitals finances. Weighing the magnitude of the affected population (5%), its financial consequences and the professionals perceived preventable impact on the worst outcome, professionals decided in favor of a more sensitive diagnostic threshold. Especially given the possible effect of undetected SGA on the chosen level of fetal surveillance during labor in the Dutch home birthing situation was considered a very undesired condition.
The Dutch obstetric system is unique with community midwives providing care for women with low risk pregnancies in primary care and clinical trained midwives, residents and gynecologists providing care for women with high-risk pregnancies in secondary and tertiary care [22]. Transfer of care from midwife to gynecologist, is a result of having one or more risk factors for pregnancy-related complications, unexpected abnormal findings and/or the occurrence of complications during pregnancy or childbirth. As a consequence, midwives view pregnancy and delivery as healthy, physiological events whereas gynecologists are more prone to be concerned about adverse events. These different views became evident during our team meetings, when professionals discussed the need and value of medical interventions.
For decisions made with help of the risk matrix, at population level, actual risk rather than risk category was directive, and an increase in extreme risk was also considered an important issue to be taken into consideration, especially as a timely instituted intervention is capable in reducing the risk. Besides risk, hierarchical structures influenced our decision-making process. Midwives experience a power imbalance in their relationships with their specialist colleagues, which can be explained from a historical perspective [23]. Midwives started as autonomous professionals, responsible for all pregnant women, only consulting another caregiver if the child had died during delivery. As a consequence, midwives were afraid of losing their autonomy by introducing new interventions and changes in current thresholds. In addition, the felt magnitude of the intervention for the individual pregnant women and their healthcare professional were valued in deciding.
We used the risk matrices in combination with the ACCORD tool, a bottom-up approach for developing collaborative interprofessional protocols. The ACCORD tool shows many similarities with the ‘Grading of Recommendations Assessment, Development, and Evaluation’ (GRADE) model and the Evidence to Decision Framework (EtDF) which are frequently used when creating a guideline. However, the main goal of the ACCORD tool is to reach consensus on content of care for clinical practice, thereby weighing current professional guidelines and protocols against the opinions and concerns of the participating disciplines [5]. Similar to the GRADE and EtDF, we identified and prioritized topics to be discussed and summarized current evidence on these topics based on existing guidelines. As guidelines represent topics on which professionals, top-down, have already agreed upon, no systematic search in medical databases was performed, as is the case with the GRADE or EtDF [24].
Obviously, the RMA approach comes with limitations, as the decisions taken are explainable but not always directly logically following the matrix. First, the design of the matrix, the number of rows and columns, and the category scaling and labels are commonly used but arbitrary. Moreover, the percentages filled in the matrix are estimates based on existing data, sensitive to changes in time and place. Therefore, there remains a certain level of subjectivity in the presented figures leaving room for individuals interpretation [2]. On the one hand, lack of exactness is unwanted, on the other hand, understanding the different aspects that are weighed during decision making helps to make balanced choices that can be explained to peers and followers. Second, although better than chance alone, a risk matrix still oversimplifies the complexity of risk. Therefore, risk matrices can only be used as a tool to support risk informed decisions, not for making or computing decisions. Assessment of likelihood, consequences and resulting risk scores actually requires subjective interpretation, whereby different users may obtain different ratings of the same quantitative risks [4]. Therefore, we have to accept that subjective decision-making will always be a part of a risk assessment process, independent of the tool used [2]. Third, risk matrices lack any time valuation, whereas the risk of a certain problem to occur in a week might be very different from the risk of the same problem in a year. A risk can be static over time while others can change overnight. For decision-making, time has to be considered as a separate factor additional to the risk matrix. Our time frame is limited to pregnancy. Time is not a concern for our studied interventions, if calcium supplementation is used in an adequate dosage of 1 g/day from 20 weeks of gestation until the end of pregnancy [25]. The risk of preeclampsia is larger in second and third trimester of pregnancy, but this is the case for every pregnant women, independent of calcium use [26]. The threshold for SGA is not time related. However, gestational age is an important factor, affecting neonatal morbidity and mortality rates, when fetal growth-restriction results in induction of labor [27]. Fourth, because it is a two-dimensional interpretation of risk I it cannot account for interventions that in themselves add a risk.
Although a risk matrix tries to overcome the uni-directional perspective of risks only to be quantified as chance by adding the impact of an adverse outcome to the decision making process, poor resolution, errors in category assignment, vague input and output may still affect resource allocation [28]. There are alternatives to the RMA that make fewer assumptions but are still not completely able to breakdown the probability and outcome as easily. These include the drug fact box and comparative effectiveness tables [29, 30]. They have advantages, and disadvantages, relative to the RMA. As we used the risk matrix in weighing and discussing possible effects of novel changes in new guidelines regarding anticipated costs and clinical effects, we decided, in an attempt to improve resolution, taking a 5 by 5 matrix instead of a less detailed matrix.