Orthopaedic oncology represents a unique subset of orthopaedic surgery both in terms of diagnoses and treatment paradigms. Decision-making has shifted from a paternalistic model to a shared decision between the physician and the patient [3, 8]. Furthermore, the manner in which the relevant information is expressed to the patient may affect their decision when choosing between limb salvage and amputation. Shared decision-making has been studied within the realm of orthopaedic surgery and is currently defined as an effort between the physician and the patient to make informed medical decisions when no “best” treatment option is available [2, 10, 11, 26]. Congruently, some medical conditions can be deemed “preference-sensitive,” and these are defined as situations in which no single treatment option is superior; as such, this applies to many orthopaedic conditions, and treatment decisions in these scenarios hinge on patients’ values, needs, and desires .
Shared decision-making entails that both the physician and the patient engage in the process. Involvement of both parties, with their different backgrounds and values, has the potential to introduce bias into this process. For example, the manner in which information is presented or “framed” to a patient may affect their ultimate decision. The effect of this framing bias has been studied in treatment decisions related to orthopaedic conditions such as tibial plateau fractures, and several other cognitive biases have been studied in the medical realm (including those dealing with the choice to place implantable cardiac defibrillators); however, we are unaware of any such studies relating to the decision between limb salvage and amputation [1, 15]. Given that multiple factors may play a role in this decision-making process, we sought to evaluate four of the main cognitive biases in this regard. Further, a patient’s inherent beliefs, values, and culture also play an integral role in their decisions regarding their healthcare. We also hoped to investigate any influence that these values or beliefs may have on a patient’s treatment decision.
We therefore asked: (1) Will the manner in which information is presented to a simulated patient, in the setting of treatment for a bone sarcoma, bias their decision regarding pursuing amputation versus limb salvage? (2) At the time of decision-making, will a simulated patient’s personal background, demographics, or mood affect their ultimate decision?
Materials and Methods
Institutional review board approval was applied for and the project was deemed exempt. A survey was designed in which a scenario was presented to a simulated patient regarding the treatment of a bone sarcoma with either amputation or limb salvage. In this study, a simulated patient was a member of the general population that participated in a blinded survey regarding bone sarcoma treatment. Survey respondents were presented with a mock clinical vignette and then asked to choose between amputation and limb salvage. Simulated patients were used in this study as it would not be feasible to study the deliberate introduction of bias into an actual patient encounter in which a choice between an amputation or limb salvage is proposed.
Experimental Overview and Survey Recruitment
Participants were recruited from Amazon’s Mechanical Turk “MTurk” platform (Seattle, WA, USA) to complete a survey hosted on Qualtrics servers (Qualtrics, Seattle, WA, USA). MTurk is a crowdsourcing platform that enables “workers” to anonymously complete tasks or surveys for monetary compensation, and has become an increasingly popular tool for research. Mturk employees, or “workers,” are generally young (median age 30), highly educated (63% hold college degrees), and female (69%) . Although it would be difficult to model a population that was identical to a sarcoma population, this population afforded the opportunity to study a population that was similar in some regards. This platform also offered distribution of the survey to a wide range of workers with varied demographic backgrounds; further, distribution of these surveys is randomized through the MTurk service and was available to its entire worker population (some exclusion criteria were applied). Recruitment in this study was restricted to individuals 18 years or older, who were classified as “MTurk Master Workers,” had previously completed > 100 Human Intelligence Tasks (HITs) on the MTurk platform with a HIT approval rating > 95%, and were geographically located in the United States (Fig. 1). Participants were excluded if they had undergone a limb salvage procedure or amputation previously.
As mentioned above, a survey was designed to present a scenario to a simulated patient (survey respondent) regarding the treatment of a bone sarcoma with either amputation or limb salvage. Specific iterations were designed to invoke cognitive bias (framing bias, anchoring bias, affect bias, bandwagon bias) by deliberate alteration of the subjective presentation of the same objective information, according to the methodology used by Bernstein et al. (Table 1) . The functional scales used within these iterations have been previously published [4, 7, 13, 14, 17, 22, 24, 25, 28-30]. Each of the four cognitive biases had an attempt, based on team experience in cognitive surveys, at a neutral language condition resulting in a total of eight iterations for the scenario (Table 1). Participants received one iteration (randomized via the MTturk platform), and completed this anonymously (Fig. 1). For example, for those allocated to framing bias, each participant completed either iteration 1 or iteration 2. A representative sample of the survey questions constructed is provided (see Appendix, Supplemental Digital Content, https://links.lww.com/CORR/A139). Overall, 404 participants completed the survey, and corresponding demographic information is also shown (Table 2).
For the framing, affect, and bandwagon biases, the primary outcome measure was the participant’s decision to undergo either limb salvage or amputation after presentation of the health scenario. The outcome measure for the anchoring bias was the highest acceptable complication rate they would be willing to tolerate to proceed with limb salvage over amputation (this was included in the survey as free response by the survey participant).
In addition to demographic information, measures were completed before and after the health scenario was presented to gauge the participant’s overall affect, depression severity, health-related quality of life, and perception of the patient-provider relationship. These included the Positive and Negative Affect Scale (PANAS), the 9-item Patient Health Questionnaire (PHQ-9), and EQ-5D, respectively [5, 6, 9, 12]. The PANAS score is a self-reported measure of affect (both positive and negative) and is scored from 10 to 50 (a lower score on negative affect indicates a more negative affect; conversely, a higher score on positive affect indicates a more positive score. The PHQ-9 is a self-administered questionnaire that is the depression module of the full Patient Health Questionnaire. It is scored on a scale of 0 to 27, with higher scores indicating more severe depression. The EQ-5D is a standardized questionnaire for measuring general health status.
The power analysis for this study was calculated based on the results of Bernstein and colleagues . Using G-Power 3.1 software (G*Power, Dusseldorf, Germany), a sample size of 48 participants per cognitive bias was required for 80% power to detect differences at a two-tailed α level of 0.05.
Analysis of the data were performed using Stata/IC 14.2 (StataCorp, College Station, TX, USA). Specifically, associations between the type of bias presented and the respondent’s choice of limb salvage versus amputation were examined. Independent-sample t-tests were used to compare means. Statistical significance was defined as p < 0.05.
The manner in which information is presented to a simulated patient was associated with that individual’s choice of amputation or limb salvage. When amputation was presented as an option to mitigate functional loss (framing bias), more patients chose amputation than when limb salvage was presented as means for increased functional gains (23 % [23 of 100] versus 10% [12 of 118], odds ratio, 2.26; 95% confidence interval, 1.07–4.77; p = 0.01) (Table 3). When exposed to affect bias, simulated patients were no more likely to choose amputation than were patients not exposed to affect bias in the manner in which the information was presented (31% [29 of 94] versus 27% [26 of 96]; OR, 0.87; 95% CI, 0.48–1.6; p = 0.657). Similarly, patients were no more likely to choose amputation when exposed to bandwagon bias than were patients not exposed to this bias (32% [25 of 79] versus 28% [30 of 109]; OR, 0.86; 95% CI, 0.47–1.50; p=0.539). The maximum risk of complication deemed acceptable by the respondents was 55% (SD +/- 26.6); when a benchmark risk of complication of 70% was presented to the respondents (anchoring bias), the maximum risk of complication acceptable increased to 60% (SD +/- 26.0, p = 0.2).
The personal background, demographics, and mood of the simulated patients surveyed all appeared to be associated with the choices they made with respect to amputation or limb salvage surgery. Older simulated patients were more likely to choose limb salvage when exposed to framing bias versus younger patients (mean age 33 years (SD +/- 7.0; 95% CI, 32–34) versus 30 years (SD +/- 10.2; 95% CI, 26–33; p = 0.02). Older simulated patients were no more likely to choose limb salvage when exposed to affect, anchoring, or bandwagon bias (mean age 33 (SD +/- 5.7; 95% CI, 31–35) versus 29 (SD +/- 10.1; 95%CI, 28.7–32.1; p = 0.07; mean age 34 (SD +/- 8.2; 95% CI, 32.3–35.7) versus 30 (SD +/- 6.7; 95% CI, 31.6–34.4; p = 0.61; mean age 35 (SD +/- 9.4; 95% CI, 33.5–36.5) versus 34 (SD +/- 11.3; 95% CI, 30.6–35.4; p = 0.5, respectively). Respondents who were employed in healthcare more commonly chose amputation versus limb salvage when exposed to framing bias (24% [eight of 35] versus 9% [17 of 183], OR, 2.46; 95% CI, 0.98–6.1; p = 0.02). This was similar when exposed to bandwagon bias (21% [11 of 55] versus 5% [six of 133]; OR, 4.43; 95% CI, 1.56–12.5; p = 0.01). When exposed to affect bias, respondents in healthcare were no more likely to choose amputation versus limb salvage (22% [12 of 55] versus 12% [16 of 135]; OR, 0.54; 95% CI, 0.24–1.22; p = 0.08). Those who chose amputation when exposed to framing, affect, and bandwagon bias, were more likely to score higher on the Pre-Survey PANAS: Negative Affect and PHQ scales (Table 4).
The treatment of bone sarcomas usually involves a surgical intervention and this surgical decision usually is a choice between limb salvage and amputation for local control of the disease. Although many factors influence this decision, including location, anatomic extent of disease, and feasibility of achieving the goals of the surgery, the choice can be a shared decision between the physician and the patient. Given that the physician educates the patient on his or her disease and treatment options, the potential for bias can be introduced that may affect the ultimate decision. This study sought to examine the effect of intentionally introduced bias into a simulated preoperative discussion on the decision between limb salvage and amputation for the treatment of bone sarcomas. We found that more people chose amputation when it was presented as a means to avoid functional loss versus when limb salvage was presented as a means to gain function. Also, simulated patients were no more likely to choose amputation when exposed to affect or bandwagon bias versus those that were not exposed to these biases. Furthermore, older patients in this study were more likely to choose limb salvage when exposed to framing bias; however, this age differential was small. Finally, if a respondent was employed in healthcare, they were more likely to choose amputation when exposed to framing and bandwagon bias.
This study had several limitations. First, this study was completed using anonymous surveys of a population of study participants. Although this afforded the opportunity to study a large cohort comprised of varied backgrounds, this could only serve as a surrogate to a decision made by an actual patient undergoing treatment for a sarcoma. There is a possibility that this population does not truly model a sarcoma population. Although this is a possibility, we used the MTurk platform due to its wide-range of “workers” (greater than 50,000 in the United States), in an attempt to capture simulated patients that could possibly substitute for a sarcoma population. However, our study population tended to be in their fourth decade of life, Caucasian, and well-educated, which may be different from the characteristics of actual sarcoma patients. Our primary goal was to study the effect on cognitive bias on decision making, and these biases have been studied in other realms of medicine (orthopaedic trauma, cardiac surgery) [1, 15]. As such, these biases also appear to be inherent to human nature; congruently, these results could possibly remain generalizable in this regard [16, 17].
Another limitation of this study also dealt with the use of simulated patients. This study would not be possible to perform in actual patients due to the extenuating circumstances inherent to the care of sarcoma patients. However, it can be argued that the gravity of a decision between limb salvage and amputation can be difficult to relay to a simulated patient. Further, the simulated patients in this study were merely survey participants without the stress anxiety that a cancer diagnosis invokes and accordingly, likely had no emotional investment into the choice between amputation or limb salvage. The results of this study may be difficult to translate to an actual clinical scenario. While it may be impossible to convey the gravity of a choice between amputation and limb salvage when not set in an actual clinical scenario, the results of this study affords an opportunity to demonstrate how bias might influence such a decision.
This model was designed specifically to test the intentional introduction of bias into the presentation of surgical options. Data were gathered pertaining mainly toward the respondents’ demographics and mood. However, other variables that are commonly involved in this treatment paradigm, such as the feasibility of limb salvage resulting from anatomic restraints or tumor characteristics, patient values (not measured in this study), or surgeon bias, were not studied. Also, we only presented options regarding limb salvage or amputation; additional options, such as the exact method of reconstruction (biologic versus metallic), or even presenting the option of expectant management as the treatment (although likely not feasible in a sarcoma population) could also introduce bias. Lastly, four specific types of biases were tested in this study. Although these biases are all relevant to shared decision-making in orthopaedic oncology, there is potential for the presence of other biases in an actual physician-patient medical decision. These factors may all confound decisions made between a physician and a patient. Although these limitations were present in our study, these were mitigated by our large number of respondents and carefully designed model to specifically examine the effect of these biases on treatment decisions.
The manner in which information is presented to a simulated patient appears to affect the decision between amputation and limb salvage in this survey setting. In these scenarios, the raw information presented to the simulated patient, in terms of functional outcomes or complications of amputation or limb salvage, was the same in both iterations [4, 7, 13, 14, 23, 24]. However, framing was used to introduce bias by presenting the same general information, but in a different manner; this forced the simulated patient to make a decision predicated on comparing the relative value of one option (limb salvage) with another (amputation). A higher percentage of respondents chose amputation when it was framed as a means to avoid functional loss as compared with the scenario in which limb salvage was presented as a means to gain function. Also, the introduction of affect and bandwagon bias did not make respondents more likely to choose amputation compared with those not exposed to these biases. It should be noted that this was a simulated scenario, and inherently, those participating in it will likely behave differently when compared with an actual clinical scenario (in which most patients wish to pursue limb salvage no matter how the information is presented). Bernstein et al.  conducted a decision-making study incorporating these same biases in the context of the treatment of different orthopaedic conditions. They were able to demonstrate a change in a decision made by patients on the introduction of these biases. Ultimately, they advocated for standardized presentations of treatment options and abstaining from presenting the choices of other patients to minimize bias . Matlock et al.  studied the prevalence of cognitive biases, and their effect on decision-making, within preoperative discussions pertaining to implantable cardioverter-defibrillators (ICDs). Across a range of cognitive biases, they frequently observed framing bias, in which the benefits of ICDs were overemphasized and the risks were minimized. Moreover, cognitive biases present during decision-making may fail to reveal a patient’s true preference. In the surgical realm, this has implications for both informed consent and satisfaction after a surgical intervention [1, 16].
We found that employment in the healthcare field, either by the respondent or a member of their family, conferred a greater degree of sensitivity to the biases introduced in these scenarios. Furthermore, a greater percentage of these respondents chose amputation as compared with those who chose limb salvage. This could be explained by a greater trust in information presented to them by a physician and, thus, an amplified response to any bias that may be introduced. However, this may also be partly explained by this study’s design, and these hypothetical decisions (in this study) could change dramatically when placed into a clinical setting. Also, knowledge of health care may help to set expectations and diminish any bias toward amputation that could theoretically be present within the general population. Quon et al.  reported that, in the setting of elective amputation, expectations were shaped by the quality of the information presented preoperatively and satisfaction was tied to these expectations.
The impact of cognitive bias appeared to be amplified in those respondents with a negative affect or mood. The proportions of depression and anxiety in populations of patients with sarcoma range from 13.7% to 33.3% and 11.8% to 47.2%, respectively [18-20, 23]. Furthermore, those patients with greater emotional distress at baseline appear to realize high rates of distress postsurgery. However, when examining psychologic outcomes of patients with sarcoma after limb salvage or amputation, outcomes appear to be similar [23, 27, 31]. Therefore, the aforementioned results indicating a greater sensitivity to cognitive bias in patients with a negative affect or mood and the potential difficulty in exposing the true preference of the patient when choosing between limb salvage and amputation have implications for postoperative psychologic outcomes.
The decision between pursing limb salvage or amputation in patients with extremity sarcoma is difficult and often reflects a complex interplay between the physician and the patient. Our study found that the intentional introduction of framing bias may affect a simulated patient’s choice for amputation versus limb salvage. Further, older simulated patients were more likely to choose limb salvage when exposed to framing bias. However, this should be interpreted with caution, as the age differential in this study was small (3 years), and the overall mean age in this study may be different than an actual sarcoma patient. Simulated patients employed in healthcare were more likely to choose amputation when exposed to this bias. Our findings may or may not pertain to the actual clinical setting of a patient with a sarcoma facing these decisions, although it is likely that biases of the physician and the patient influence decision-making. Future studies may include investigations of the effects of cognitive bias on the physician, or on the influence of other factors on this process, such as standardized decision-making or communication aids. These findings support the present knowledge of physicians who understand that how they present facts and opinions to a given patient facing a difficult treatment decision likely shapes their decision-making process.
1. Bernstein J, Kupperman E, Kandel LA, Ahn J. Shared decision making, fast and slow: implications for informed consent, resource utilization, and patient satisfaction in orthopedic surgery. J Am Acad Orthop Surg.
2. Bozic KJ. Orthopaedic healthcare worldwide: shared medical decision making in orthopaedics. Clin Orthop Relat Res.
3. Charles C, Gafni A, Whelan T. Shared decision-making in the medical encounter: what does it mean? (or it takes at least two to tango). Soc Sci Med. 1997;44:681–692.
4. Clayer M, Doyle S, Sangha N, Grimer R. The Toronto extremity salvage score in unoperated controls: an age, gender, and country comparison. Sarcoma. 2012;2012:717213.
5. Crawford JR, Henry JD. The Positive and Negative Affect Schedule (PANAS): construct validity, measurement properties and normative data in a large non-clinical sample. Br J Clin Psychol. 2004;43:245–265.
6. Cvengros JA, Christensen AJ, Cunningham C, Hillis SL, Kaboli PJ. Patient preference for and reports of provider behavior: impact of symmetry on patient outcomes. Health Psychol. 2009;28:660-667.
7. Davis AM, Devlin M, Griffin AM, Wunder JS, Bell RS. Functional outcome in amputation versus limb sparing of patients with lower extremity sarcoma: a matched case-control study. Arch Phys Med Rehabil. 1999;80:615–618.
8. Emanuel EJ, Emanuel LL. Four models of the physician-patient relationship. JAMA. 1992;267:2221–2226.
9. Herdman M, Gudex C, Lloyd A, Janssen MF, Kind P, Parkin D, Bonsel G, Badia X. Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L). Qual Life Res. 2011;20:1727-1736.
10. Informed Medical Decisions Foundation. Informed medical decisions. Available at: http://www.informedmedicaldecisions.org/what-is-shared-decision-making/shared-decision-making-resources/
. Accessed July 15, 2018.
11. Klifto K, Klifto C, Slover J. Current concepts of shared decision making in orthopedic surgery. Cur Rev Musculoskelet Med.
12. Kroenke K, Spitzer RL, Williams JBW. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med.
13. Li J, Wang Z, Ji C, Chen G, Liu D, Zhu H. What are the oncologic and function outcomes after joint salvage resections for juxtaarticular osteosarcoma about the knee? Clin Orthop Relat Res. 2017;475:2095-2104.
14. Malek F, Somerson JS, Mitchel S, Williams RP. Does limb-salvage surgery offer patients better quality of life and functional capacity than amputation? Clin Orthop Relat Res.
15. Matlock DD, Jones J, Nowels CT, Jenkins A, Allen LA, Kutner JS. Evidence of cognitive bias in decision making around implantable-cardioverter defibrillators: a qualitive framework analysis. J Card Fail.
16. Mazur DJ, Hickam DH. Treatment preference of patient and physicians: influences of summary data when framing effects are controlled. Med Decis Making. 1990;10:2-5.
17. Mishra S, Gregson M, Lalumiere ML. Framing effects and risk-sensitive decision making. Br J Psychol. 2012;103:83-97.
18. Pardasaney PK, Sullivan PE, Portney LF, Mankin HJ. Advantage of limb salvage over amputation for proximal lower extremity lower extremity tumors. Clin Orthop Relat Res. 2006;444:201-208.
19. Paredes T, Canavarro MC, Simoes MR. Anxiety and depression in sarcoma patients: emotional adjustment and its determinants in the different phases of disease. Eur J Oncol Nurs. 2011;15:73-79.
20. Paredes T, Periera M, Simoes MR, Canavarro MC. A longitudinal study on emotional adjustment of sarcoma patients: the determinant role of demographic, clinical and coping variables. Eur J Cancer Care (Engl). 2012;21:41-51.
21. Quon DL, Dudek NL, Marks M, Boutet M, Varpio L. A qualitive study of factors influencing the decision to have an elective amputation. J Bone Joint Surg Am. 2011;92:2087-2092.
22. Redelmeier DA, Rozin P, Kahneman D. Understanding patients' decisions cognitive and emotional perspectives. JAMA.
23. Refaat Y, Gunnoe J, Hornicek FJ, Mankin HJ. Comparison of quality of life after amputation or limb salvage. Clin Orthop Relat Res. 2002;397:298-305.
24. Rougraff BT, Simon MA, Kneisl JS, Greenberg DB, Mankin HJ. Limb salvage compared with amputation for osteosarcoma of the distal end of the femur. A long-term oncological, functional, and quality-of-life study. J Bone Joint Surg Am. 1994;76:649-656.
25. Sherif M. The Psychology of Social Norms. New York, NY: Octagon Books; 1965.
26. Slover J, Shue J, Koenig K. Shared decision-making in orthopaedic surgery. Clin Orthop Relat Res. 2012;470:1046-1053.
27. Tang MH, Castle DJ, Choong PF. Identifying the prevalence, trajectory, and determinants of psychological distress in extremity sarcoma. Sarcoma. 2015;2015:745163.
28. Tversky A, Kahneman D. Judgments under uncertainty: heuristics and biases. Science. 1974;18:1124-1131.
29. Tversky A, Kahneman D: The framing of decisions and the psychology of choice. Science. 1981;211:453-458.
30. Van Egmond-van Dam JC, Bekkering WP, Bramer JAM, Beishuizen A, Fiocco M, Dijkstra PDS. Functional outcome after surgery in patients with bone sarcoma around the knee; results form a long-term prospective study. J Surg Oncol. 2017;115:1028-1032.
31. Weddington WW, Seagraves KB, Simon MA. Psychological outcome of extremity sarcoma survivors undergoing amputation or limb salvage. J Clin Oncol. 1985;3:1393-1399.
32. Zaldivar A, Ross J, Irani L, Tomlinson B. Who are the crowdworkers? Shifting demographics in Amazon mechanical turk. CHI EA. 2010:2863-2872.