Several factors influence decision making, and understanding these factors helps in understanding the decision making process and potential outcomes. Because medical decisions have an impact on patients’ health, they should be of high quality (Klein, 2005). However, medical practice is complex and time is often limited, leading to use of shortcuts to decisions, or heuristics. Unfortunately, the use of heuristics also brings with it some pitfalls (Klein, 2005). Despite being highly trained, doctors are prone to making mistakes (Bornstein and Emler, 2001); cognitive biases may detract from the use of logical and statistical decision heuristics (Hershberger et al., 1994). This research will address the theoretical aspects of decision making before considering factors that influence decision making in the context of clinical practice. Further, it will address heuristics of medical problem solving and decision making. Finally, a summary will be offered.
2.4.1 Theoretical Framework
Decision making is important in medical practice, and because health outcomes are probabilistic, most decisions are made under conditions of uncertainty (Kaplan and Frosch, 2005). Medical science has not yet solved the uncertainty surrounding many medical decisions (Gillett, 2004) and is the least developed aspect of evidence-based practice (Spring, 2008). Furthermore, uncertainty has been characterized as the most important factor influencing physician behaviour (Arrow, 1963). Thus, uncertainty is central to Decision Theory concerned with medical decision making.
Decision theories are developed to explain decisions under three main conditions of consequence of choice: certainty, risk and uncertainty. Decision under certainty applies when all decisions will lead to only one consequence. Decision under risk applies when a choice will have one of several possible consequences and the probabilities of the consequences are known (Heylighen, 2010). Contrary to decision under risk, decision under uncertainty applies when a choice will have one of several possible consequences, but the probabilities of the consequences are not known. Furthermore, decision theories fall into two main categories:
normative or prescriptive theories are based on idealized situations where a decision can be fully rational and all probabilities calculated; and descriptive theories work on the principle that people do not necessarily follow axioms and thus describe what people do rather than focus on optimality. Evidence-based medicine is concerned with integrating individual clinical expertise and the best
external evidence (Sackett et al., 1996), and works on the basis of known probabilities (Kaplan and Frosch, 2005). However, in a knowledge-for-critical evaluation project, Greenfield et al. (2007) address the usefulness of the results of randomized controlled trials for clinical and policy application. The authors conclude that the evidence includes patients who may have minimal benefit and that generalization to patients excluded from the study may result in overtreatment. Thus, it may be argued that the axioms of evidence-based medicine may be invalid in individual treatment decisions, and consequently, normative decision theories explaining decisions under risk may be not be applicable in medical decision making. Further support is provided by (Eddy, 1994) in that clinicians often disagree about appropriate action in similar clinical situations. In addition, there is evidence to support large geographical variations in the use of health care (Wennberg et al., 2002, 2004) and that this variation does not result in the same variation in patient outcomes (Fisher et al., 2003, 2003). This leaves descriptive theories as a possible framework for medical decision making. However, evidence based theories are not abundant and the lack thereof to support medical decision making has led to clinical approaches and decision tools being based on assumptions - and these assumptions have been challenged by behavioural research (Reyna and Rivers, 2008). Modern theories supported by empirical evidence differ in their views of risky decision making, behavioural change, health promotion and medical decision making (Reyna and Rivers, 2008). Three such theories have recently been gaining recognition: theory of reasoned action and its extension the Theory of Planned Behaviour, the trans-theoretical model and fuzzy trace theory.
2.4.1.1 Theory of Reasoned Action and the Theory of Planned Behaviour The theory of reasoned action states that intent is the best indicator of motivation to act (Fishbein, 2008) and is concerned with what determines intention. Intention to perform a certain action is a function of two factors: attitude and subjective norm (perception of importance) (Fishbein, 2008). Determinants of attitude have been identified as: evaluation of belief and strength of belief (O'Keefe, 2002) p.103-104. Perceived lack of ability may have an impact on intention to perform an action, and in 1991, Ajzen revised the theory and in so doing, added a third factor, perceived behavioural control, leading to the Theory of Planned Behaviour (Ajzen, 1991).
2.4.1.2 The Trans-Theoretical Model
The trans-theoretical model is concerned with strategies or process for change and originally contained four stages of change; pre-contemplation, contemplation, action and maintenance (Prochaska and DiClemente, 1987). The work of Prochaska and DiClemente (1987) was based on the need for behaviour change models having a bearing on health, with special focus on addiction. This change model has led to goal setting as an accepted method to improve performance and has been operationalized through practice guidelines within the context of medical decision making (Prochaska, 2008).
2.4.1.3 Fuzzy Trace Theory and Adaptive Decisions Making
The fuzzy trace theory is based on the finding that people rely on the gist of information, its bottom line meaning, as opposed to verbatim details in judgement and decision making (Reyna, 2008). Thus, fuzzy trace theory lends itself to explaining why detailed information about risk does not necessarily support medical decisions. However, fuzzy trace theory is at odds with the central assumption of evidence-based medicine, that decisions made by computation are superior to those made by intuition (Spring, 2008). Important aspects of decision need not be conscious, and subjective perception of reality shapes decision making (Reyna and Rivers, 2008). Decision making by medical experts often relies on intuitive gist processing and pattern recognition (Lorenz et al., 2005). However, decision tasks vary in complexity and the limited processing capabilities of the decision maker make the adoption of adaptive decision strategies imperative. As no one theory will suffice to cover all eventualities (Reyna and Rivers, 2008), the work of (Payne et al., 1993) still stands as a credible seminal work having greatly contributed to the understanding of adaptive decision processes. The framework for adaptive decision behaviour distinguishes between decision strategies and problem solving; decision strategies are generally reserved for diffuse problems and problem solving lends itself more to an hypothetico-deductive approach (Payne et al., 1993) p.60. This concept is translated to clinical practice by (Elstein and Schwarz, 2002). Furthermore, the adaptive decision framework relies on trade-offs between cognitive effort and accuracy. Thus, the fundamentals of adaptive decision making are that individuals decide how to decide.
2.4.1.4 Heuristics
Human decision behaviour is a contingent form of information processing, and different decision strategies may be used to suit the situation (Payne et al., 1993) p.9. Heuristics may be considered mental short cuts reducing decision effort (Shah and Oppenheimer, 2008). Focus on decision effort follows the assumption that cognitive effort is limited (Payne et al., 1993) p.73. However, decision effort is not the only part of the heuristic; it is also important to consider the accuracy of the strategy to yield a good decision, and hence the effort-accuracy framework was posited by (Payne et al., 1993) p.72. Unfortunately, the use of heuristics also brings with it some pitfalls (Klein, 2005, Tversky and Kahneman, 1974). Despite being highly trained, doctors are prone to making mistakes (Bornstein and Emler, 2001) and cognitive biases may detract from the use of logical and statistical decision heuristics (Hershberger et al., 1994).
In a seminal paper, Tversky and Kahneman (1974) showed that people rely on a limited number of heuristic principles, reducing the complexity of decision making.
The authors describe heuristics and biases employed for making decisions under uncertainty. According to Tversky and Kahneman (1974), the main heuristics employed when making decisions under uncertainty are: representativeness, availability and adjustment & anchoring. Klein (2005) identified five pitfalls in decisions about prescribing: representativeness heuristic, availability heuristic, overconfidence, confirmatory bias and illusory correlation. These pitfalls have
anything but a rejuvenation of the debate set in a prescription context. However, the notion is further supported by (Poses and Anthony, 1991), but the authors use the term ‘inappropriate use of heuristics’. Furthermore, Payne et al. (1993) pp.193-216 identified the lack of knowledge and execution of heuristics as a pitfall. Thus, is can be argued that identifying an appropriate decision strategy and executing within the given context are central to successful decision making.
The representativeness heuristic (Kahneman and Tversky, 1979) is an assumption that something seemingly similar to other things in a category is a part of that category (Klein, 2005). It is important to be aware of base rate occurrence and not weighting information inappropriately. The availability heuristic builds on the principle of the ease with which information can be accessed and gives too much weight to information that is easily available (Payne et al., 1993) pp. 201-207. In many situations, estimates are made from a starting point - and this is known as anchoring {Tversky, 1974 #534@@author-year}. If necessary adjustments are not made, then anchoring bias may occur as a decision is based on incomplete reasoning due to lack of baseline information.
Even though heuristics are available, inappropriate use may lead to mistakes (Poses and Anthony, 1991), and it is important to consider why this may be the case. Confidence relies on representativeness of the prediction with disregard for predictive accuracy (Tversky and Kahneman, 1974). According to Klein (2005), most people are more confident about judgments than they ought to be and may be contextually influenced by an individual’s attitude, leading to overconfidence or illusion of validity(Tversky and Kahneman, 1974, Crano and Prislin, 2006).
According to Tversky and Kahneman (1974), illusory correlation is a pitfall of the availability heuristic. How frequently two events occur together will influence the associative bond between them (Payne et al., 1993, Tversky and Kahneman, 1974). Klein (2005) defines this simply as: “Illusory correlation is the tendency to perceive two events as causally related, when in fact the connection between them is coincidental or even non-existent.”
Tversky and Kahneman (1974) argue that adjustment and anchoring bias may occur as a consequence of subjective probability distribution and hence violate the logic of statistical prediction. The fact that people do not appear to follow logic and statistical theory of prediction leads to intuitive judgement (Poses and Anthony, 1991) that sometimes leads to severe and systematic errors (Kahneman and Tversky, 1973).
In conclusion, decision under uncertainty is based on the use of a limited number of heuristics designed to reduce the effort of judgement, but this may reduce the accuracy of the decision and lead to errors (Kahneman and Tversky, 1973, Tversky and Kahneman, 1974, Poses and Anthony, 1991, Payne et al., 1993, Hershberger et al., 1994).
2.4.2 Medical Decision Making and Prescribing Behaviour
From the literature, it is evident that complex decisions are based on the use of a limited set of heuristics to reduce the effort associated with decision making.
Medical decisions are designed to have effects on patient health, and hence the
quality of these decisions is important. Supported by the positivist ontology of evidence based medicine, most medical decisions are probabilistic in nature and occur under uncertainty (Elstein and Schwarz, 2002). Given the potential high risk to patient health and complex nature of modern medicine, factors affecting the quality of medical decisions need to be understood. Medicines play an important role in modern medicine (Hemminki, 1975) and their use as a therapeutic intervention is widespread.
The work of (Bradley, 1991) represents the only identified comprehensive literature review on decision making and prescribing behaviour linking the observed effects and moderators highlighted thus far to the theory of decision making. At first glance, it may seem that the author adopts a neutral stance legitimate for a knowledge-for-understanding project, but adopts a negative stance criticising the rash introduction of interventions to curb cost without the understanding of how decisions are affected. In short, Bradley (1991) concludes that the understanding of prescribing behaviour requires study of the underlying decision process. This author argues that policy change may have untoward effects on patients if it is made without an understanding of current decision patterns. This sentiment is later echoed by others (Sketris et al., 2009), further underlining the need for more research on the topic.
Taking a broader view, McKinlay et al. (1996) found that the variability in medical decision making by physicians was not entirely accounted for by prescriptive theory and thus conclude that non-medical factors such as patient characteristics, physician characteristics and practice setting may play an important role in influencing physician prescribing behaviour. The notion of the authors is supported in a more recent literature review by (Hajjaj et al., 2010) and by the work of Gill et al. (1999), reporting that only 51 % of interventions introduced to change physician prescribing behaviour had an effect, and that no difference could be found across interventions. Finally, Bornstein and Emler (2001) found that physician clinical reasoning is vulnerable to biases and that formal processes may aid in improving decision quality by focusing on the most relevant information. The overall finding is that how factors influence physician decision making is a phenomenon poorly understood. In conclusion, more studies may lead to better decisions and healthier patients (Poses and Anthony, 1991).
In an interesting study, employing a grounded theory approach, on the decision-making process leading to appendectomy, Larsson et al. (2004) developed a model suggesting that the decision is made based on the interplay between an assessment of the patient’s condition and contextual characteristics. The authors did not strictly follow Glaser’s advice on theoretical sampling, thus limiting the depth of findings from the study and consequent conclusion. However, the study indicates that contextual factors may have a bearing on medical decisions, fully in line with the previous discussion on agency and agency relationships.
According to Miller (1989), decision control involves direct influence over outcomes, but the range of outcomes are typically preselected; on the other hand, process control involves indirect influence over outcomes. Process for deciding
and the individual physician has only indirect influence over process outcomes.
The process will guide the decision making, but intuitive judgements are still made (Kahneman and Tversky, 1973, Poses and Anthony, 1991) within the bounds of decision and process control.
2.4.3 Shared Decision Making
Shared decision making is being advocated as the preferred method for decision making in medical care. In an intellectual project for knowledge generation, (Charles et al., 1997) identified four characteristics of shared decision-making in the medical encounter: both physician and patient must be involved, both parties share information, consensus regarding preferred treatment must be built, and that agreement on implementation is reached. The approach outlined by Charles et al. (1997) will reduce the informational asymmetry between patient and physician and is denoted by others as the informed treatment decision model (Gafni et al., 1998). In an attempt to ascertain difference in patient outcome between the informed treatment decision model and the physician as a perfect agent for the patient, Gafni et al. (1998) investigated adjuvant chemotherapy versus no adjuvant chemotherapy in patients with early stage breast cancer. The authors concluded that both approaches result in the same outcome. However, the distinction between the two agency approaches is delegation (perfect physician agency) versus retention of authority (informed treatment decision model) (Gafni et al., 1998). Nevertheless, even though shared decision making models are advocated, implementation remains limited in practice (Barratt, 2008).
2.4.4 Summary
Decision making in medicine is performed under uncertainty. However, no general theory of medical decision has been formulated, but the theory of reasoned action and its extension the Theory of Planned Behaviour is the most studied theoretical framework informing on the topic of physician cognition leading to decision behaviour. In addition, Agency Theory plays a central role in defining contextual contributors to the decision making process. Despite being highly trained, doctors are prone to making mistakes, and cognitive biases may detract from the use of logical and statistical decision heuristics. Given the multidimensional theoretical framework having a bearing on medical decision making, deciding how to decide is central.