Shared decision making: Does a physician's decision‐making style affect patient participation in treatment choices for primary immunodeficiency?

Abstract Overall health care spending in the United States is equivalent to more than 15% of GDP, yet outcomes rank below the top 25 in most quality categories when compared with other Organization for Economic Cooperation and Development (OECD) countries. The majority of spending is consumed by small patient populations with chronic diseases. Experts believe increased patient‐physician shared decision making (SDM) should result in better overall longitudinal care but understanding the physician's role in facilitating SDM is limited. Structural equation modelling was applied to results of a 2016 questionnaire‐based survey of 330 US physicians who treat approximately 55% of primary immune deficiency requiring immune globulin therapy; it tested the relationship between slow/rational vs fast/intuitive decision‐making styles and SDM as mediated by patient‐centric care and moderated by physician's trust in the patient. The results showed a statistically significant relationship between slow/rational decision making and SDM. The results also suggest differences related to age, gender, education, and race but no differences related to trust.

incurred. 4,5 To alleviate this quality and cost gap, the Institute of Medicine (IOM) has recommended the implementation of patientcentric goals and targeting those patients with chronic conditions. 6,7 Therefore, understanding the relationship between physician decision making and patient participation for chronic disease management may provide insight into the factors that facilitate implementation of patient-centric goals and improve care.
Chronic diseases are conditions that last more than 3 months and often involve complex interacting comorbidities that interact in difficult to predict and emergent ways. 8,9 The complexity and need for long-term treatment of chronic conditions is one of the reasons why half of US health care costs are spent on 5% of patients. 10 Primary immunodeficiency (PID) was chosen to represent chronic diseases. PID is a set of rare genetic disorders that prevent the immune system from functioning normally [11][12][13][14][15] and is one of the most expensive diseases to treat due to cost of drug, preventable hospitalizations, and difficulty in diagnosis. 16 Physicians treating PID with immune globulin therapy were sought to participate in a study to understand factors that influence their decision-making process.
One approach to care that meets the IOM recommendation and has potential to improve optimize patient outcomes of chronic diseases is that of shared decision making (SDM). [17][18][19] SDM is a process whereby health care providers and their patients make treatment decisions jointly 20 ; it is defined as an "approach where clinicians and patients share the best available evidence when faced with the task of making decisions, and where patients are supported to consider options, to achieve informed preferences." 21 SDM is considered the "pinnacle" of patient care. 18 However, an understanding of the implementation of SDM with chronic diseases and how doctors actually interact with patients is limited. 22,23 Furthermore, factors that influence physician clinical decision making and SDM adoption have not been examined and tested within the context of specific chronic diseases. 24 The need for additional research and the lack of an integrated SDM theory are an identified gap in the literature and if developed will contribute to enhancing the use and impact of SDM in medical decision making. 25,26 Recently proposed SDM models could benefit from an understanding of these factors. [27][28][29][30] Dual process theory integrates two forms of thinking and, as such, can be applied to the study of clinical processes and decision making.
Heuristic decision making (System 1) is an approach that relies on experience. 31,32 Physicians simplify information by forming standard approaches to treatment-based clinical experience, which have been termed "illness scripts." 33,34 Rational decision making (DMR) (System 2) is the slower approach, one that requires effort and conscious analysis. 32,35 Physicians may analyse different factors, such as the ratio of harm to benefits, especially when there is no clear or standard procedure given unique or complex patient circumstances. 36 Although a physician makes use of either system according to circumstances, it has been shown that individuals have an affinity for one over the other. 37 Eva and Croskerry have reported differences regarding the role of heuristic and "deliberate analytical intervention." 38,39 Although they pointed out that both rational and heuristic methods are deployed, often in iteration, much of this discussion has been in the context of diagnostic and acute care rather than patient participation and chronic care. 39 Furthermore, a physician's "patient-centric biopsychosocial approach" should mediate or explain why SDM is adopted. 40,41 To explore the relationships between decision-making styles and other physician characteristics on SDM, we compiled a survey questionnaire based on published validated scales and used structural equation modelling to measure how these characteristics were interrelated. The hypotheses to be explored were as follows: (a) Do physician decision-making styles affect patient participation in SDM as mediated by patient-centric care? (b) Does the level of trust between physicians and patients influence the effect of physician decision-making style on SDM? and (c) Do physician traits such as age, gender, education, and race influence the effect of their decision-making style on SDM?

| STUDY DESIGN AND METHODS
This study was a quantitative analysis of physician decision making in the treatment of PID. A web-based survey was completed by 330 US physicians who use immune globulin therapy to treat PID to look for possible links between physician decision-making style and the use of patient-physician SDM.

| Constructs of interest and their measurement
Sixty questions were used in the survey to characterize decisionmaking style, patient-centric approach, and trust in patients. Each item score was set to a 5-point Likert-type scale ranging from 1 = "strongly disagree" to 5 = "strongly agree." The questions are shown in Tables  A physician's patient-centric approach was measured using the questions from the Patient-Practitioner Orientation Scale (PPOS). 43 These include questions to patients on how they feel. Physician's trust in patients to provide accurate information was measured using the 12 questions from Thom's 2011 Physician Trust in Patient Scale. 44 Two aspects of patient participation in decision making were measured. The first was participation in the choice of treatment schedules APC_4 I tend to…encourage patients to ask a lot of questions they have discussed with other patients.

APC_5
I tend to…think of patients as equals in the treatment process.

APC_6
I tend to…match treatment to fit a patient's lifestyle.

APC_7
I tend to…learn the patients' culture and background.

APC_8
I tend to…use humour with my patients.

APC_9
I tend to…focus treatment decisions on patient preferences.

DMR_10
I tend to…learn as much as I can about possible consequences before making decisions.

DMR_2
I tend to…rarely make a decision without gathering all the information I can find.

DMR_4
I tend to…only make treatment decisions when all the information is gathered and available.

DMR_7
I tend to…substantially rely on published clinical data for treatment decisions.

DMR_8
I tend to…make decisions slowly to ensure I make the right decisions.

DMR_9
I tend to…see each of my decisions as stages toward a definite goal.

IOP_1
Patients tend to…inform me of challenges with treatment schedules.

IOP_10
Patients tend to…convince me to modify the protocol based on their input.

IOP_5
Patients tend to…play a key role in organizing a treatment plan.

IOP_6
Patients tend to…decide how the treatment will be administered (e.g. subcutaneous or intravenous).

IOP_7r
[reversed] Patients tend NOT to…accept the protocols I suggest as-is.

IOP_8
Patients tend to…participate in setting new protocols for treatment.

IOP_9
Patients tend to…influence my decision regarding treatment protocols with their opinions. (Continues)

IOT_2
Patients tend to…get better results if they are on medication that they requested.

IOT_3
Patients tend to…request medications they've read about in advertisements.

IOT_4
Patients tend to…request medications they've heard about in social media.

IOT_5
Patients tend to…request medications they've heard about from other patients.

IOT_6
Patients tend to…feel free to voice their product preference during our meetings.

IOT_7
Patients tend to…freely make comments on the treatment product.

IOT_8
Patients tend to…share feedback on the products I recommend to treat their condition.

IOT_9
Patients tend to…actively participate in the product choice to treat their condition.

TIP_1
I trust the patient will…provide accurate medical information.
TIP_2 I trust the patient will…let me know when there has been a major change in their condition.

TIP_3
I trust the patient will…tell me about all medications they are using.

TIP_4
I trust the patient will…follow the treatment plan exactly as I have provided.

TIP_5
I trust the patient will…manage their condition with the prescribed treatment plan.

TIP_6
I trust the patient will…tell me if they are not following the treatment plan.

TIP_7
I trust the patient will…not manipulate the office visit for secondary gain. and administration methods (described hereafter as participation in treatment protocols). This was measured using the Gallan, Jarvis participation scale 45 and Siegel and Ruh's scale. 46 The second was participation as would occur in choice of medication or device (described hereafter as participation in treatment tools

| Statistical analysis
Exploratory and confirmatory factor analyses (EFA and CFA) are standard multivariate statistical techniques used to develop a theory or model. 48,49 EFAs explore how data relate to variables (factors) and were performed serially to conduct tests to confirm or increase the validity of the item set. CFA is used to confirm or reject the theory or model and was conducted based on the EFA results. The results of the CFA indicate that all survey items load significantly (P < .001) and substantively (standardized factor loadings .7 or above) on their respective theoretical constructs, supporting convergent validity and reliability as shown in Table S1. Convergent validity between constructs means there are expected to be related and are, in fact, related. 50 Methods bias was tested for by using a chi-squared difference test between the unconstrained common method factor model and the fully constrained zero common method factor model 51,52 and with a latent marker variable method that tested the chi-squared difference between nested models (unconstrained, equal, and zero). 53 A factor likely to have an effect was the physician's age; the older and more experienced immunologists had a distinguishable demeanour to care than the younger sample. Therefore, a new model was constructed, which included age and excluded trust. The EFA and CFA remained sufficient after removing "trust" items, as suggested by Table S2. Both analyses were conducted with data from the survey using SPSS (version 23) and AMOS software. Details of the procedures are included in the online Appendix.

| STUDY RESULTS AND MODEL
A total of 350 physicians completed the online survey; 20 responses were excluded from data analysis; therefore, the sample group consists of 330 completed surveys (the sample group). The most common reason for exclusion was outlier characteristics (as predefined Cook's distance test score of greater than .05), and these characteristics seemed to be mostly due to an unengaged response with the participant giving identical scores to all items in a construct. The low Participants were 74% male, and 72% were White Tables 3.
Twenty-one percent of the participants were in the specialty of allergy/immunology.
The results showed a statistically significant relationship between slow/DMR and SDM. The initial path model was constructed from the 17 survey items remaining after the EFA and CFA. See Figure S1.
There are multiple significant pathways from DMR to patient participation with protocols and tools. However, pathways that were not significant were trimmed from the model to achieve model fit Furthermore, DMR has a strong relationship with patient-centric approach (β = .37, P < .01), which had a mediating influence on the relationships between DMR and patient participation with protocols (IOP) and tools (IOT). This suggests that physicians incorporating DMR seldom include patient participation with tools decisions but encourage participation with protocols decisions.

| Multi-group
A statistical test (chi-square difference test) was performed to determine whether physician characteristics recorded in the survey affect the level of patient participation. The results also suggest differences related to race, age, education, and gender but no differences related to trust. These differences suggest that with White physicians, the use of heuristic decision making increases patient participation with treatment tools (β = .264, P < .001), but is not the case for non-White physicians (β = .088, P = .363). Additionally, increasing age increases patient participation for white physicians whereas age does not affect participation for non-White physicians. PhD-degreed physicians encourage patient participation and SDM by extension, with or without a patient-centric approach. Additionally, age increases patient participation for PhD-degreed physicians (β = .367, P = .055) when choosing treatment protocols.  63 We acknowledge that the present study only surveyed physicians, not patients nor other members of the care team. It is probable that other insights may be obtainable with the use of patient interviews and questionnaire items and constructs for measuring the level of patient empowerment/enablement.
The current study was necessarily limited by the questionnaire items and constructs (instruments) used. An underlying assumption was that responses would identify physicians' decision-making styles in a binary manner. However, it is equally true that individual physicians alternate between decision-making styles according to circumstances. 35  The study was also limited by its cross-sectional design using a survey taken one time only. For chronic care with treatments administered over the long-term, this may not capture the full influence of factors affecting patient participation. Including a longitudinal study in future research would be useful in this regard.
Lastly, we realize that PID is a rare disease; therefore, the findings and conclusions from this study cannot be generalized to other chronic conditions. Future studies in other more common chronic diseases are warranted.

| Policy implications for the management of chronic disease and SDM
"Let's give the patients the choice" is a frequent mantra in chronic patient populations such as PID. 68 Often cited models of care promote patient choice in treatment protocols and tools although practical implementation has been difficult. 69 These models assume rational actors that can efficiently and optimally implement SDM through thoughtful deliberation and balanced collaboration. Such models also assume SDM is a technique that can be taught algorithmically despite enormous clinical complexity and far reaching innovative and expensive therapeutic advances. Perhaps a better starting point is physician self-assessment of decision-making styles and potential biases with the aim to develop personality interventions that enhance SDM. 37 In the study described herein, a systematic method was used to look for associations between physician decision-making styles and demographic characteristics and the use of SDM in the treatment of a model chronic disease. Statistically significant factors were identified which health care providers should consider when developing SDM treatment models. For example, DMR may require more physician time; workloads, compensation, and time constraints issues should be considered. 70 Likewise, despite some claims, chronic care management is often complex and incompletely understood; therefore, innovative therapies may require more experience and education, consistent with adaptive health practices. 29,71 The role of trust should be better understood since its presence is thought to be essential to SDM. 72 interactions always as simple as two-way discussions, given that other members of a care team can also be important in establishing trust?
The study found that decision-making style, whether intuitive or rational, is associated with the level of patient participation and suggests that perhaps some traditional assumptions underlying SDM should be reassessed and more research focus should be on physician behaviour as opposed to patient behaviour. 76,77 Or, at a minimum, several physician-driven factors should be considered when designing optimal physician-patient interaction.
The current study was limited to PID, and it is possible that the fac-