**2. Theoretical Background**

The clinical decision process is a complex dynamic and under pressure process including a choice between options as categories and diagnosis (Hausmann et al. 2016; Higgs et al. 2019); its complexity is due to the involvement of more people and the gap between information availability and those necessary. Hence, clinicians, often have to face uncertainty and make decisions without definitive information, taking into account the so-called "imperfect information" (Higgs et al. 2019, p. 504) that is available at the early stage of the clinical encounter (Cooper and Frain 2016). At this stage of the decision process, physicians can be satisfied only in deciding what kind of information has to be collected and which aspects of the situation have to be pointed out (Higgs et al. 2019). Thus, they are able to provide only a potential diagnosis.

In this context, the use of big data surely improves clinical output (Atoum and Al-Jarallah 2019); in particular, data availability (Sun and Scanlon 2019) would represent an information source which is able to sustain the decision-making process, by making it more aware (and bias-free) as based on concrete findings drawn from similar clinical situations (Yan et al. 2017).

Within the clinical reasoning sphere, however, it is crucial to refer to a personal understanding of the patient's condition by the physician and to his/her ability to make a decision. Within this path, indeed, the physician has to evaluate several elements concerning the patient history (e.g., findings from clinical examination, test results) and, above all, they have to make a decision under time pressure that can considerably condition the whole process (Cooper and Frain 2016; Goldsby et al. 2020). According to Del Mar et al. (2006) "*doctors have to be good at interpreting, at prioritising, at making compromises, at seeing what matters*" (p. vi), with the aim to managing complications and health crisis in a timely way; in a complex environment, in fact, it could often happen that "*what seems right in theory would be damaging* *in the flesh*" Del Mar et al. (2006, p. vi). In particular, from Zavala et al. (2018) we learned that clinical decision-making can be influenced by factors which increase complexity and uncertainty depending on the specific case; thus, factors which complicate the healthcare context can be summarised as the following: unpredictable workflows, non-replicable conditions, pressures, organisational systems, workload, teamwork, human interactions, and patient complexity.

Moreover, it should also be noted that within the clinical decision process the state and the action spaces of physician (decision-maker) are strictly influenced by the patient. There is a sort of interdependence between the two actors that can lead the entire process and its results (Lippa et al. 2017). In particular, more is the clinician's self-e fficacy in her/his patient managemen<sup>t</sup> abilities more is the patient reliance on that clinician's approach (Sizer et al. 2016).

In addition, the clinical decision process could be influenced by several factors like external context (Robinson et al. 2020), environment, the complexity of the task (Higgs et al. 2019) and by the capabilities, confidence, and emotions of the practitioner (Smith et al. 2007). Actually, it is impossible to analyse the clinical decision process without considering the context or situation in which it occurs, as reported by several authors (Cooper and Frain 2016; McBee et al. 2015; Fargen and Friedman 2014) context characteristics or interactions between physician, patient, and environment are fundamental to understanding the whole process of clinical reasoning and can modify it in several ways, sometimes bringing errors.

Hughes and Nimmo (Cooper and Frain 2016) in particular have identified five types of errors that can occur in the diagnostic process and are related to:


According to Simon, there are several restrictions on human cognition related to the social environment in which decisions occur; in the author's opinion, an individual has a "bounded rationality" and thus people make "satisficing rather than optimal decisions" (Cristofaro 2017b, p. 172).

Moreover, according to Kahneman (2011), there are two modes or systems of thinking:


By using system one, people incur cognitive shortcuts that can a ffect the decision-making process (Cristofaro 2017a). Biases are very common in every human situation, so in clinical practice too; knowledge and experience cannot avoid the possibility to make these errors: they are "subconscious deviations in judgement leading to perceptual distortion, inaccurate judgement and illogical interpretation", in Cooper's opinion (Cooper and Frain 2016, p. 26); these errors can be related to both systems of thinking.

To help decision-makers to avoid these biases and improve the decision-making process, Kahneman et al. (2011) identified the checklist: a tool to improve the quality of decisions finding defects in the process. This tool provides a set of 12 questions aimed at identifying errors in thinking (biases). To use this tool, a third person is required, independent from the analysed group; indeed, people cannot recognise their own errors as a third person (Kahneman et al. 2011).

Specifically, in the healthcare field, (Antonacci et al. 2020) identifies a twofold direction between (i) biases in emergency care; (ii) biases in clinical and therapy medicine.

Emergency physicians are required to make decisions under an extremely high level of uncertainty, and they have to consider plenty of factors (Croskerry 2003). Indeed, during their decision-making process in an emergency room or similar situation, physicians have to consider not only the physical

patient condition (choosing which treatment to exclude, which one has to be initiated and when) but also patients' preferences, resources availability, cost and time (in particular time and resources are limited in this context) (Hausmann et al. 2016). Another aspect that is important to underline is that in the emergency decision-making process, clinicians have to choose how to allocate their time and e ffort and which patient to prioritise. This factor along with limited knowledge of the patient's personal history by the physician makes the emergency room a "natural laboratory of error" (Croskerry 2003; Antonacci et al. 2020; Hausmann et al. 2016). According to Croskerry (2003, p. 776), moreover, "nowhere in medicine is rationality more bounded by relatively poor access to information and with limited time to process it". Accordingly, emergency medicine is an area full of heuristics (Abatecola 2014): a method of solving problems by finding practical ways of dealing with them, learning from past experience (Oxford Dictionary 2012).

Regarding the clinical decision process related to therapy and clinical medicine, there are several studies that can explain which kind of factors may lead this process in di fferent fields; listed below there are some sample factors that can modify the decision-making process in some medicine fields:


Furthermore, medicine literature has identified and described more than one hundred cognitive biases (Cohen and Burgin 2016), listed below there are some of the major biases of other medicine fields:


Moreover, some scholars recognised the ownership of a Hospital Institution as able to directly/indirectly influence the clinical decision process in surgery. The main di fference between public and private hospitals in the surgical field regards the type of healthcare intervention provided to users/patients; public-access hospitals carry out more traumatic and emergency interventions on acute patients; private hospitals, instead, tend to provide mostly elective and planned surgery (Ierano et al. 2019).

As a consequence, another significant di fference between public and private hospitals is about physician autonomy in decision-making. Accordingly, Ierano et al. (2019) stated that "autonomy was perceived to be greater in the private hospital setting than in the public hospital setting". In their study, the interviewed nursing sta ff noted that private surgeons had the capability to "dictate their own practice" irrespective of the guidelines or the hospital policy, as the private physicians were "doing their own thing and renting the space".

Concerning the field of this study, orthopaedic surgery, it is important to underline that it focuses on both the emergency and clinical/therapy medicine fields; therefore, the orthopaedic surgery field can be led both by "emergency bias" (mostly heuristics) and by the "clinical biases", depending on which kind of healthcare services we focus on. Particularly, Sizer et al. (2016) proposed a model to drive the clinical decision process in the orthopaedic field: the "evidence-supported practice wheel" that poses the clinician's expertise and the patient at the centre of the problem, it makes the physician more flexible to adapt to the patient's needs and context, still relying on scientific literature. According to the author, some technical factors influence decision-making in orthopaedic surgery; Sizer defines the biomedical information on the patient that has to be taken by the physician in order to provide an aware decision-making process.

In addition, as stated by Grove et al. (2015), orthopaedic surgery is characterised by high professionalisation based on long-training and proven practical experience. In this field, according to the author, an elite group of surgeons (usually grouped per di fferent countries and regions on the base of specialist surgery) is recognised as the reference key-opinion leaders able to influence (ordinary) surgeons in decision-makings through their researches and case study reports developed in their working experience. In our opinion, these circumstances could a ffect the orthopaedist's decision-making with cognitive biases (as identified by Kahneman), which might display the guidelines from the key-opinion leaders as always valid, without any further in-depth consideration of the patient case.

Moreover, some scholars belonging to the industrial field state that the working context and job experience are also able to influence the decision-making sphere (Hendrick 1999; Kobus et al. 2001). Accordingly, the expectation of this manuscript, for the healthcare field, is to understand how patient information together with physicians' variables would impact personal thinking as a base of the decision-making process.

### **3. Materials and Methods**

Qualitative research (Patton 2002) was a better fit for the types of study that we conduct and for the state of prior research and theory we refer to (Edmondson and Mcmanus 2007). Accordingly, qualitative descriptions allow the researchers to stand by the data and provide factual summaries of participants' experiences and perceptions (Neergaard et al. 2009). Given the above, to analyse di fferent scenarios of the orthopaedic surgery world, three semi-structured "face to face" (El Said 2017) interviews with three di fferent orthopaedic surgeons from di fferent working contexts in Italy were conducted. In particular:


The respondents work in Italy, in di fferent regions.

Even if only three, in the authors' opinion, these interviewees represent a quite good depiction of the orthopaedic landscape in Italy (Torre et al. 2017); the Italian healthcare, according to Spano and Aroni (2018) is based on public-access hospitals/health authorities (which mostly provide free of charge services as responses to emergencies and scheduled surgeries) and private organisations (which exclusively sell scheduled services).

Accordingly, considering respondents belonging to di fferent ownerships (public and private) of healthcare organisations with di fferent roles and experience means estimating the main variables concerning the physician job: working experience and working context. Particularly, the respondents' sample (even if small) considers:


Please, note that all the three interviewees have the responsibility of their working team, and they are the main coordinators of the decision-making during their specific work-shifts and surgical activities. Accordingly, they could be considered as key-expert-informants (Yin 2004, 2017) for the aim of this study, given also its explorative-goal (Scapens 1990) based on critical case study sampling (Patton 2002) design. This explains why it would not make sense to interview a law-experienced physician, working in a private context, whose contribution to the decision-making process would be surely considered as secondary.

Hence, in order to contextualise and define the topic, a definition of follow-up was required by all respondents. Thus, to achieve this study goal, Kahneman's checklist (as the quality control tool aimed at monitoring causes and e ffects of cognitive distortions) was submitted to the three respondents; on the basis of their expertise, interviewees were asked to highlight and explain those biases recognised on their current decision-making, with specific reference on hospital discharges (Jette et al. 2003).

Kahneman's checklist aims, in fact, to find those biases related to the possibility that decisions can be distorted by cognitive mistakes/biases.

Thus, according to Cristofaro (2017a), the checklist was adjusted to make it suitable for the healthcare context and in particular it was focused on the intellectual process regarding the decision-making about patient's follow-up (Jette et al. 2003) after knee or hip arthroplasty.

Table 1 reports the adjusted checklist submitted to interviewees, modified to be suitable for this study's aims. The last column of Table 1 contains the link between each checklist question and the related control question(s), separately listed at the end of the table.


**Table 1.** Adjusted checklist and biases which it refers to.


**Table 1.** *Cont.*

Source: Authors' elaboration inspired by Kahneman et al. (2011), Cristofaro (2017a), Stylianou (2008). The \* it is referred to the text in the twelfth line. For this reason, the text in that line begins with \*.

As stated by Stylianou (2008), a control question is defined as a probe question "that controls an independent variable in the participant's thinking for verification and exploration purposes" (p. 242). Particularly, control questions should be used in situations in which "the substantive theme contains multidimensional concepts or complex causal structures" (Stylianou 2008, p. 242) that need to be disentangled.

As in our case study, control questions allow the interviewer to monitor that the questions protocols were respected and that all the inquiry issues have been understood by interviewees in coherence with the study goals.

Thus, as included in the previous Table 1, the control questions were provided for the following reasons, concerning the understandability of the qualitative results detected:


Precisely, orthopaedics participated in video-conference semi-structured interviews conducted by the principal investigator between April and June 2020. Each interview lasted between 60 and 80 min and was digitally recorded. The analytic process was guided by the principles of conventional content analysis (Hsieh and Shannon 2005), the interviews after the recording were transcribed verbatim, identifying information was removed, and data were stored in a password-protected computer. Verbatim transcription was investigated through the thematic analysis approach (Braun and Clarke 2006), following the methodological fit drawn from (Edmondson and

Mcmanus 2007). The results of the ongoing analysis were reviewed by the authors together with interviewees directly during the regular meetings; disagreements were resolved by discussion and consensus.

Proceeding further, Figure 1 shows the phases carried out in the methodology. Specifically, it includes the chronological order of all steps followed and the inclusion/exclusion criteria considered for interviewees choice.

**Figure 1.** Step of analysis. Source: Author's elaboration.
