Are Physicians' Decisions Affected by Multiple Nonclinical Factors?

Physicians' decision-making during patient encounters is multifactorial and complex. Decisions may be affected by multiple nonclinical factors but the scope of these influences and their potential effects remain unclear. Qualitative interviews with clinicians were analyzed and complemented by MEDLINE search. Fifteen clinicians raised many undue nonclinical factors agreeing that they were often operative and relatively recent. Altogether, 75 nonclinical factors and barriers that may adversely influence the quality and objectivity of clinical decisions today were identified. Many were highly prevalent. They were grouped into 4 major domains: outside forces (n=13); components of the encounter (n=22); physician's personal and cognitive factors (n=22); and patient-related factors acting on the physician (n=18). A significant impact on the quality of care, resource utilization and patient-physician relationship is suggested by the interviews and literature. Unwarranted practice variation which is ubiquitous may also be related to nonclinical factors. Most research is limited, based on physicians' surveys and response to vignettes. Alternative prospective methods are suggested. Thus, decision-making by physicians appears to be often affected by multiple, ubiquitous, and potentially inappropriate nonclinical factors. Meanwhile, multifaceted educational efforts and system changes are feasible and likely to reduce the potential untoward effects which may be substantial.


Introduction
Choices and judgments are the essence of clinical life and physicians are called upon to make the right decision countless times a day. Clinical decision-making is a challenging science and art. The medical literature is quite clear on how decisions should be made: clinical methods are used to carefully collect all the facts and identify a pertinent question. This is followed by scrutinizing the current literature and applying clinical judgment in selecting the best suited path, explaining it, discussing options with the patient, then committing the decision to paper or screen. However, this ideal approach, a 'hallmark of professional competence' [1] which is deliberate and thoughtful, complete and systematic, evidence-based, patient-centered, utterly objective and essentially pure of alien interfering factors -may be actually practiced far less often than we care to think.
Medical decision-making remains a complex process integrating many variables. Professional excellence and the patient's welfare are not the only concerns that drive clinical decisions. It has long been recognized that in reality, physicians' behavior, choices and decisions during the encounter are susceptible to many nonclinical pressures and influences [2]. Multiple studies revealed that essential practice patterns and decisions may be affected by patients', physicians' or organizational characteristics. Notable examples include communication [3][4][5]; test-ordering, prescribing, or referrals [6][7][8][9][10][11][12][13][14]; procedures [14][15][16][17][18][19][20]; admissions [21,22] and post-hospitalization care [23][24][25]. Practice variations that are not driven by clinical indications are highly important because they occur across a very large spectrum of conditions and may lead to significant adverse patient health outcomes [26][27][28][29][30][31][32][33] which may be preventable. Nevertheless, a comprehensive view of all such factors is lacking in the literature and is increasingly relevant since many recent changes may have added to the problem. We have undertaken this study to identify, map and classify potential nonclinical factors that may adversely affect physicians' decisions in their patient encounters today.

Methods
The study was conducted at an academic medical center in Israel where the health system is basically similar to that in the US: primary care and ambulatory consultations are delivered through Health Maintenance Organizations (HMO) and referrals to regional hospitals are done as needed. The author conducted in-depth qualitative interviews (40 -140 min.; mean 80 min.) with a systematically-drawn sample of physicians adhering to RATS criteria (Relevance of study question; Appropriateness of Method; Transparency of procedure; Soundness of interpretation). These criteria have been suggested for ensuring that quality requirements for quality research are met (Clark JP, http://www.biomedcentral.com/authors/rats). Experienced (≥15 years since graduation) clinicians (about half in primary care, half hospitalists) were arbitrarily selected (i.e. No. 10, 20, etc.) from alphabetical lists of hospitalists (internal medicine and its subspecialties) or primary care physicians in central Israel. A long clinical experience was deemed mandatory to ensure a thorough recognition of all aspects of the patient-physician encounter and to be able to evaluate possible changes with time. Physicians were contacted by phone and an interview arranged following verbal agreement. If the physician selected did not meet experience criteria or declined, the next one was contacted. The study was approved by our Institutional Review Board (0051-08-KMC). The basic structure of the interview is given in Appendix I. Participants were encouraged to speak freely and support their responses with actual examples. Dialogues were recorded and transcripts analyzed to create a detailed list of nonclinical pressures and influences affecting physicians' decisions. Since the behavioral science literature suggests that individuals have difficulty identifying and articulating many highly-important influences on decision-making, it was decided that data should be complemented by a literature review. Themes identified in the interviews served as key words in the following literature search. Additional key search terms were derived from relevant manuscripts identified by the original, interviews-based search. The main goal was to map as many nonclinical factors affecting physicians as possible, thus a systematic review was not warranted. Original research articles in English cited in MEDLINE (1.1991 -12.2011) were searched. Typical Medical Subject Headings (MeSH) terms used included couplets of: physician demographics (hospitalists, primary care physicians) OR practice characteristics (HMO, private/group practice), with physicians' behavior(attitude of health personnel, physician-patient relations) OR physician's decisions (decision making, physician's practice patterns, defensive medicine) OR quality of health care (outcome assessment, health services misuse, healthcare disparities, unnecessary procedures, medical errors). Relevant articles were retrieved and read and their reference lists searched for further relevant research. We included studies dealing with 'external' factors affecting communication or decisions and outcome studies identifying unwarranted variation in healthcare -defined as variation not explained by illness, patient preference or evidence-based medicine (EBM). Manuscripts that were not original research articles (such as opinion articles and reviews) were excluded. Analysis included identification and listing of any nonclinical factor affecting physicians which was found to be significant.

Results
The clinicians selected constitute a roughly representative sample of clinicians in central Israel (Table 1). None declined the interviews (100%). Examples of clinicians' insights brought up in the interviews concerning nonclinical considerations affecting their daily patient encounters are cited in Appendix II. The nonclinical factors that came up most often were time shortage and its aftermaths (14/15); untoward influences on prescribing or referrals (11/15); and responses to clinical uncertainty / risk of litigation (9/15). All clinicians 'strongly agreed' (13/15) or 'agreed' (2/15) that many nonclinical factors often influenced their decisions. Nine believed that more than half of the nonclinical factors affecting their decisions were new (i.e. "not active when they started to practice") and six estimated that about half of them were new (Appendix I). No physician in our sample thought that just 'some' or 'none' of the factors were new.   *A few of the nonclinical factors show overlap with clinical ones or may be classified in more than one domain. For example, the patient's age can be regarded both as a clinical as well as nonclinical factor; the need to demonstrate improved practice or hospital statistics can be an 'outside force' as well as a physician's personal factor.
"Outside forces" are factors extraneous to the encounter and common to all patients seen, such as directives of the health care system. Altogether, 13 different 'outside forces' were identified (Table  3).   @ Frequently missing pertinent clinical information @ Impaired continuity of care @ Fragmentation of care between many physicians of different specialties @ Impaired communication between primary physicians and specialists / hospitalists @ * Outstanding sensitivity of current tests yielding many false positive results mandating attention and evaluation @ * Frequent, often complex problems beyond the physician's field of expertise @ * The need to practice evidence-based medicine at the point of care @ * Problems related to data in the literature (Information overload, absent pertinent data, variable interpretation of data, mismatch vs. the individual patient, 'care gaps') @ * Overflow of diagnostic and therapeutic options to choose from (Tests, drugs, etc.) @ * Continuing expansion of guidelines from multiple professional societies, presenting extensive demands, sometimes unrealistic or conflicting @ * Complexity of communication about choices (Multiple options, shared decisions) @ * Obligatory 'new' contents of the encounter (Patient's health literacy; shared decisions, prevention and early detection measures) @ Increase in physician's administrative duties (e.g. getting approval for tests or drugs) @ Computer and electronic health record (EHR) constraints @ Quality of data presentation in the patient's medical record (hard to find, indecipherable, etc.) Table 4: Components of the encounter (n=22) *An originally 'clinical' factor may become 'nonclinical' when excessively time-consuming and demanding, pressuring clinicians to compromise and opt for sub-optimal, potentially inappropriate choices.
"Physician's personal or cognitive factors" are personal and psychological characteristics influencing decisions and behavior. Altogether, 22 different 'physician's personal or cognitive factors' were identified (Table 5).
@ Clinical uncertainty -fear of error, failure, losing esteem @ Fear of litigation or complaint -assurance or avoidance behavior @ Physician's acute fatigue (effect of queue position and time of day on performance) @ Physician's burnout, anxiety and stress @ Physician's job satisfaction, when low @ Responses to errors that occurred (Personal and professional) @ Competence concerns, impaired self-efficacy (Confidence in ability to achieve task) @ Fear of losing the patient's trust or of damaging the relationship @ Recoil from intimacy, emotional involvement or suffering @ Physician-related contextual factors (age, sex, yrs. since graduation, religion, ethnicity, background) @ Physician's specialty -Tendency to "pull" cases toward specialty and use its tools @ Heuristics and biases # in test ordering, diagnostic reasoning or treatment selection @ Physician's attitudes towards risk ('risk preference') @ A tendency to prefer New tests and drugs, regardless of incomplete data, unproven safety and expense @ Inertia -lack of incentives to change old practices @ Personal disagreement with accepted authoritative guidelines @ Judgment about other fellow professionals influencing referrals and response to recommendations @ Compassionate optimism -favoring a 'better' diagnosis than justified out of compassion to the patient @ Physician's overconfidence @ Lack of effective feedback about patients' outcomes @ Physician's personal convenience considerations @ Physician's financial incentives, both direct ('self-referral') and indirect  # beyond participatory decision-making, taking into account the patient's values and preferences as essential part of a patient-centered approach and the patient's autonomy. * Pressure for referral may be particularly common [37] but pressure for tests and drugs also applies.
Examples of physicians' comments from the transcripts of the interviews are given in Appendix II, arranged according to the four domains that we identified. An examination of the wide array of factors reveals several possible classifications of nonclinical factors that may affect physicians' decisions are described in Table 7.
Not all listed factors are 'inappropriate', unwarranted or potentially deleterious. For example, patients' demands or pressures may be clinically unfounded, but taking them into account is part of patientcentered care which is at the core of good practice [38]. Managers' constraints may not always be in the best interest of the individual patient, but they serve to contain escalating system expenditures. Some primarily 'clinical' factors are listed because they may exert an adverse effect on decisions in the context of real-life time constraints. For example, increasing patients' age, comorbidities and polypharmacy; the need to practice EBM at the point of care; or to educate the patient and achieve participatory decision-making. Another distinction is between factors that are inevitable and factors that are modifiable. However, even in the case of non-modifiable factors, physicians' response to them may be amenable to improvement so that the ultimate quality of care would not suffer. New nonclinical factors (such as computer and EHR constraints or managers' directives)  (Tables 3-6) are cumulative and not mutually exclusive. Both research methods strongly suggest that they often act on physicians in varying combinations with the potential to deviate optimal physicians' behaviour and objective, evidence-based decisionmaking towards improvised shortcuts, suboptimal choices and questionable actions. Such a plethora of 'interfering' factors in each of the domains was found and associated with considerable prevalence and significance that it was surprising to discover that the subject -in its entirety -remains understudied. We did not find a single current study addressing the full spectrum of nonclinical factors and their effect on medical care. Most studies found revolve around physician's responses to surveys/ clinical scenarios or around observations of outcomes in populations demonstrating variations in care or redundant testing. Direct prospective 'real time' observations of physicians' performance and prospective studies were rare. A novel research method to try and overcome this difficulty is suggested in Appendix III.

Discussion
The qualitative interviews identified a large variety of nonclinical factors that play a significant role in physicians' practice patterns and

What changes brought them forth?
The escalating costs of medical care and ascendance of managed care have added a third party into the once intimate equation of patient and physician and introduced conflicts of interest. Business managers are now closely monitoring physicians and pressuring them to see more patients, request less (and less-expensive) tests, prescribe cheaper medications and shorten hospital stays [56]. The impact on the doctor-patient relationship and physicians' independence and job satisfaction is considerable [57]. The information and communication technology revolution exposed physicians to supervision by managers; made patients expect physicians to be always available and provided direct immediate access to hundreds of items of detailed data on each patient as well as to an unlimited number of articles and guidelines [58]. However, overburdened schedules and growing time pressures often prevent their proper utilization [59]. A recent study in three countries reveals that time constraints are a widespread significant problem, as are physicians' feelings of increasing burden and diminishing control [52]. The unprecedented advances in biological insight and medical high-technology are yielding sophisticated capabilities and increasing options [60] -but patients are also older, more complex and overmedicated [61]. The remarkable sensitivity of modern imaging yielding numerous 'incidental' findings of dubious significance and the persistent rates of misdiagnosis, attest to the fact that uncertainty in medicine has not diminished and may even have increased [62] deeply affecting decision-making [50]. The notorious 'defensive medicine' is one undesirable response promoting much redundant (and far from harmless) testing, imaging and referrals to protect against litigation [12,49]. It may also involve avoidance of 'problem' patients or 'risky' (but indicated) procedures. The recognition of the patient's autonomy and the need of better informing patients, improving their health-literacy, assessing their values and preferences, and aiming at shared-decisions [63] have added substantially to physicians' tasks with no additional time provided [64]. Patients nowadays are also more knowledgeable, demanding and critical and patient pressure is often perceived by physicians [65]. This consumerist manner and increasing competition, not least by complementary and alternative medicine providers, contribute to physicians' burden [66,67] and the high prevalence of burnout common to many settings (12). Today's multicultural society is also adding new difficulties to clinical encounters [68]. Finally, the computer has become an additional partner in the patient-physician relationship. Along with its benefits, the electronic health record may absorb the physician's eyes and attention, dictate the tempo of the encounter and influence its content (by presenting prompts requiring attention) while detracting from the contact with the patient [69].

Overall impact on decisions
Clearly, these changes in medical care are profound, powerful and positive [70]. Although they have been often discussed separately, their inevitable overall impact on the quality of physicians' decisionmaking, patient-physician relationship and health outcomes had only partially and imperfectly been investigated. Nevertheless, given the ubiquitous presence, diversity and large number of nonclinical factors found (Tables 3-6) the quality and objectivity of physicians' decisionmaking should be questioned since it may often be seriously jeopardized. For example, the consequences of working under time constraints alone (cited by 14/15 clinicians) include decreased quality of decisions [71], avoidance of essential tasks [9,59,72], decreased physician job satisfaction and high stress, burnout and fatigue triggering further deterioration in empathy, decisions and quality of care [73]. As a result, patient-physician relationship may deteriorate. On the patients' side, unmet expectations lead to dissatisfaction, decreased trust, compromised adherence, increased symptom burden, increased utilization of health services and worse outcomes [74][75][76].
Almost all clinicians interviewed 'strongly agreed' that their daily decisions were often affected by many diverse nonclinical factors (Appendix II). This is in agreement with the literature cited (see Introduction). A prominent effect of nonclinical factors on a large variety of decisions was often demonstrated [42][43][44][45][46][47][48], as well as an association with adverse patient outcomes [26][27][28][29][30][31][32][33] in hospital care and even more in ambulatory settings [77,78]. Thus, nonclinical factors have been associated with impaired physician performance including superficial history and examination; failure to search resources for evidence-based answers; redundant activities; suboptimal treatment choices; deficient provision of preventive services; impaired participatory decision-making [79] and poor attention to patients' concerns and suffering [80,81]. Serious and widespread problems of underuse, overuse and misuse of healthcare ensue [12,82] with compromised quality of care and potential patient harm [12,65,75,83]. Nonclinical factors contribute to significant disparities in health care that prevent certain patients from getting the benefits of evidencebased treatments [42][43][44][45][46][47][48]. They also represent a salient reason for the common ordering of excessive and frequently redundant laboratory tests [7,35,84]; inappropriate diagnostic imaging which is often associated with patient harm [85,86]; unnecessary prescriptions and referrals [9,12,65]; and poorly-indicated follow up [49]. Thus, the factors identified may culminate in a marked detrimental effect on any variable of importance in medicine including 'hard' biological outcomes; patient wellness; resource utilization and provider's satisfaction.

Current research deficiencies and future objectives
One of our findings was that most research on the effects of nonclinical factors on decisions fell short of revealing the full scope of the problem. Much as it was thirty years ago [2], the putative link between nonclinical factors, skewed decisions and adverse patient outcomes remains incompletely substantiated. Most research found was indirect and predominantly based on surrogate markers such as physicians' responses to surveys or clinical scenarios and on documenting suboptimal care for certain subgroups of patients. However, responses to written case simulations do not necessarily measure actual clinical behaviour [87,88]. Limitations of surveys abound [89,90] and physicians' judgment of their own performance is notoriously over-rated [91]. Research demonstrating unwarranted variations of care are more robust, but these studies are selective (focused on cardiac procedures, joint replacement surgery, etc.) and study specific populations vs. controls (women, elderly, blacks, low SES, etc.). Moreover, variations in care are complex to decipher and may have alternative explanations. Finally, these studies judge decisions indirectly, after the fact, by looking at their cumulative results rather than analyzing them in real-time. Only very few studies have prospectively examined actual physician behaviour with patients [4,65,92]. Studies sequentially examining the occurrence of biased physicians' decisions as part of the whole spectrum of their actual daily clinical work were conspicuously absent. Such studies are more difficult to perform but not unfeasible (Appendix III) [93,94]. Further research is mandatory and likely to lead to improved physicians' coping, minimizing the negative impact of nonclinical factors.

Can decisions be improved?
Some contextual factors deviating physicians' decisions are inherent (such as clinical uncertainty or patient /physician /setting characteristics), but many are modifiable [58,86,[95][96][97][98][99][100]. Even when factors are non-modifiable, physicians' responses to them can be improved (e.g. coping with uncertainty, time management, etc.). For a successful intervention, several principles apply. First, interventions have to be multifaceted; and involve both system changes and educational efforts. Second, they must start early -preferably at medical school [97] but could also commence later [100]. Third, optimally, they should be continued indefinitely. Implementing advances in technology may allow physicians rapid online access to the entire patient's health information across institutions as well as to databases such as Up-to-date that may support patient-tailored decisions within a short time frame [101]. Additional system changes to determine allocation of sufficient time per patient to accommodate new tasks may be required [102], particularly in the ambulatory setting [64]. Meanwhile, agenda-setting and prioritization [103] and improved utilization of available time can be acquired by training, to achieve high quality communication [96,97], quality examination [104], rational test-ordering [86,105], data-based decisions [58,95] and patient-centered approach [98][99][100]. Many errors and biased decisions can be traced to failure in these tasks [32,74,[106][107][108][109]. Workloads that are poorly tolerated are also a cause of errors, burnout and inappropriate actions [28,29,32,73,110]. As mentioned, time management that helps in dealing with workloads can be improved by learning [96][97][98]. A brief training can make students significantly more likely to appreciate and attend to contextual patient factors [111]. Activities that promote physician personal awareness improve their personal reactions in encounters [112]. A CME program that focused on self-awareness had achieved improved physician well-being, better protection against burnout and increased likelihood of providing patient-centered care [100]. Small-group meetings to discuss problem patients or stressful experiences may improve physicians' coping [113,114]. Interventions promoting clinical and humanitarian skills [81,115] and enhancing the dissemination of methods for skilful information management, data evaluation, problem solving and decision-making are needed. Physicians' susceptibility to cognitive biases can also be improved with focused instruction [116,117]. Becoming cognizant of the many potential pitfalls on the way to optimal decision-making (Tables 3-6) can also enable a more reflective type of practice and improved physicians' coping.

Study limitations
Clinical care varies between countries, health care systems, ambulatory or hospitalized patients and primary or specialist care. Nevertheless, the Israeli health system is similar to other Western countries and many of the factors identified seem universal and common to clinical encounters in all settings implying that generalizability is possible. We have elected to address nonclinical factors in any setting as one entity, whereas, inpatient care is quite different from primary care or ambulatory care within hospitals or specialty clinics. Nevertheless, physicians' decisions are facing basically identical pressures and barriers regardless of the setting, and thus, a broader view is advantageous. Only 15 clinicians were interviewed, but qualitative studies based on in-depth interviews have often been small [8] and our adherence to recommended strategies for qualitative research [118] assured rigor and validity. The format used may have resulted in 'recall bias' which could not be eliminated. However, the many recurring themes found seemed independent of any particular memorable cases and supported by the literature. Finally, the strategy of the literature review generally followed the Cochrane Collaboration criteria [119] but meta-analysis was not performed. Since our goal was the compilation of a comprehensive list of contextual interfering factors influencing decisions ("mapping") this search strategy is adequate and does not undermine the conclusions.

Conclusions
The rapid advances in medicine are associated with an increasing presence of nonclinical factors, barriers and pressures adversely affecting physician's decisions, the quality of care and the patientphysician relationship. Dozens of nonclinical factors can be identified and most can be termed 'inappropriate' and are highly prevalent and widely distributed regardless of setting. These factors may have a significant negative effect on physicians' judgment and choices. As a result, the proportion of patient-physician encounters in which physicians' actions are pure, unbiased and evidence-based; patientcentered and comprehensive may be considerably more meager than believed. The total effect on patient outcomes may be substantial but remains unknown, awaiting future quality studies.
Decision-making by humans can never be entirely objective. Some of the barriers discussed are inevitable and ingrained in today's medical practice. However, much can be done to improve physicians' coping and adjustment to the changing environment. Educational interventions may increase physicians' personal awareness and decrease the potentially detrimental effect of nonclinical factors. This may improve the application of evidence-based medicine, the objectivity of decisions and the quality of care. Only then will we be able to reap the outstanding benefits of modern medicine.