From William Osler’s attributed assertion that “the good physician treats the disease; the great physician treats the patient who has the disease” up to present writings on social medicine, there is a widespread appreciation that “applying social science principles to medicine—a practice sometimes called ‘social medicine’—enables us to contextualize patient care to achieve more sustainable and equitable health outcomes.”1
But what, exactly, is the great physician doing to contextualize patient care? Specifically, can we distill those skills into a set of observable behaviors? And, finally, are such behaviors in fact predictive of better health care outcomes? In an era of performance measurement and value-based care, these are questions that physician educators, payers, and providers themselves all should be asking.
Physicians are now widely assessed for their adherence to guidelines which are, in turn, based on research evidence.2 However, that research evidence is intrinsically devoid of patient context. While a particular medication may outperform placebo in a randomized controlled trial, it may be of no benefit to subgroups of patients who can’t afford it, don’t understand how to take it, or have competing responsibilities that preclude adherence to the studied dosing schedule.3 The physician who is treating the patient with the disease rather than just the disease will accommodate these challenges. What is lacking is a performance measure for physician attention to such contextual factors. In short, while we assess physicians in their performance at following guidelines, we are not assessing whether they know when not to, or when following them is not enough, and what to do instead.
How Does Inattention to Patient Context Lead to Medical Errors?
What are the implications of omitting patient context in care planning? In 2004, one of us explored the question through a case analysis of a middle-aged woman with obesity, referred for bariatric surgery after unsuccessful attempts at more conservative measures to reduce her weight.4 She had two complications of overweight: diabetes and hypertension, and a history of adhesions following a cholesystectomy. The research evidence supported the decision to recommend surgery from a risk–benefit standpoint, despite the need for an open rather than laparoscopic procedure; and her stated preference was to have the surgery done.
What had not been explored, however, was her life context. The physician picked up on an offhand comment she made that one reason she wanted the surgery was that she would be “better able to take care of her son.” When the doctor inquired about what was wrong with her son, she poured out how she had sole responsibility for lifting, bathing, and feeding a young man with end-stage muscular dystrophy, supporting her eight-year-old daughter, and tolerating an abusive, alcoholic husband who she’d not thrown out because she needed the money that came from his disability income and a small pension he received. When the physician observed that she might not be able to lift the boy for weeks after the surgery because of the risk of wound dehiscence, the direction of the conversation and care plan changed. After more discussion, the patient concluded that this was the wrong time for her to have the surgery. She canceled the procedure. An error in planning was averted.
What do we call this sort of error? The Institute of Medicine has defined a medical error as “the failure of a planned action to be completed as intended or the use of a wrong plan to achieve an aim.”5 Sending the patient to the operating room at that particular time would have been “the wrong plan to achieve an aim”—her aim, which was to care for her son. It would also have put her at unnecessary risk of a postsurgical complication as noted above. It would have been a medical error.
Unlike other kinds of medical errors, however, it would not have been possible to detect from a review of the medical record because the information needed to know that the care plan was wrong—information about the patient’s context—would not have been recorded if it were overlooked. The case illustrates how these particular types of medical errors, which occur because of an inattention to patient context, fly under the radar of chart-based approaches to detecting medical error.6 We decided to assign them a descriptive name. We called them “contextual errors.”7 A contextual error is what happens when the clinician is treating the disease rather than the patient who has the disease.
Contextual errors occur when clinicians overlook patient context that is essential to planning appropriate care.7,8 We define patient context as all that is expressed outside the boundaries of a patient’s skin that is relevant to planning the patient’s care. Patient context may be organized into at least 10 broad domains including competing responsibilities, social support, access to care, financial situation, and skills and abilities, among others (Table 1).4 Note that it includes emotional state and spiritual beliefs that—while originating in the brain—are manifest as behaviors and actions expressed “outside the skin” and, hence, are a part of the context of a patient’s care.
How common are cases like the one we’ve described, in which inattention to patient context results in a contextual error? How does one detect them? How often do they occur? Are they preventable? And, perhaps most important, do they really matter, in terms of their impact on health care outcomes and costs? We describe here a series of studies we have conducted to explore these questions. We conclude by considering the implications of what we’ve learned for further study, and offer some preliminary recommendations that may reduce contextual error rates through performance improvement interventions and medical education reforms.
How Is Attention to Patient Context Evaluated?
Recognizing that we are not the first to identify inattention to social/contextual factors as compromising care, we canvassed the literature for specific tools and strategies for identifying contextual errors in care delivery and did not find any. There has, of course, been much consideration of the importance of psychosocial factors in the process of care planning, dating to the work of George Engel, followed by many initiatives to incorporate psychosocial and biomedical elements into patient care.9–16 What we were unable to identify, however was a method for ascertaining whether relevant psychosocial factors have, in fact, been integrated into a care plan. Available instruments for analyzing physician–patient communication behavior do not track this specific task.17
We conceptualize the process of “contextualizing care,” as we term it, as consisting of addressing four sequential questions during a clinical encounter18: first, are there clues—“contextual red flags”—suggesting that context essential to care planning may be present? In our case study, the patient’s comment about her son is a contextual red flag. Second, if a contextual red flag is present, did the clinician explore it? We term this a “contextual probe.” For this illustration, that entailed asking about her son. Third, are there, in fact, “contextual factors” essential to care planning, revealed either in response to probing or unsolicited? The patient’s responsibility for lifting and bathing her son is a contextual factor because addressing it in the care plan is necessary to avoiding contextual error. Finally, was the contextual factor addressed in a contextualized care plan? That meant canceling the surgery following a discussion of the patient’s situation. Table 1 provides examples of contextual red flags, probes, factors, and contextualized care plans across 10 domains of context.
Some might argue that a physician should not rely on contextual red flags but, rather, should take a more comprehensive approach to taking a social history in which all patients are routinely asked about potential life challenges such as caretaker responsibilities, inability to pay, or loss of social support. We regard such an approach as impractical and akin to the exhaustive review of systems taught to second-year medical student but soon replaced by a hypothesis-driven approach. Not all care requires contextualization. An insured woman with an uncomplicated urinary tract infection may simply need a prescription for an antibiotic. Routinely inquiring whether such a patient can afford the medication is impractical. On the other hand, if she hints that she may not take the antibiotics (a contextual red flag), and happens to be pregnant, the clinician should probe for possible contextual factors in the domains of “skills and abilities” (a lack of knowledge about the consequences) and cultural perspective/spiritual beliefs (favoring traditional remedies), among others.
As outlined above, knowing whether context is a factor in care planning and, if so, whether those factors were addressed, requires hearing if clues are present, if the right questions have been asked, if the patient in turn revealed contextual factors essential to care planning, and if those factors were subsequently accommodated. Hence, knowing whether care is appropriately contextualized or whether there are contextual errors requires listening in on the visit.
Do Physicians Make Contextual Errors?
To determine whether physicians make contextual errors when given an opportunity to do so, and to benchmark the rate of these errors against errors that occur when physicians overlook biomedical information, from April 2007 to April 2009 we sent unannounced standardized patients (USPs) to internal medicine ambulatory practices across the Chicago, Illinois, and Milwaukee, Wisconsin, areas. They presented with one of four clinical scenarios, each with four different variants of common conditions, such as asthma, in which appropriate care depended on attention either to a contextual factor (e.g., can’t afford medication), a biomedical factor (e.g., untreated gastroesophageal reflux), both, or neither (i.e., on simply following guidelines; termed an “uncomplicated” variant). Our team documented, based on audio recordings of the encounters, the length of the visits and whether the physicians identified the complicating factors by probing red flags (biomedical or contextual), and addressed them.19,20
Several striking findings emerged from an analysis of 399 visits20: First, whereas physicians provided error-free care in 73% of uncomplicated encounters, their care was appropriate in only 38% of biomedically complex encounters, 22% of contextually complex encounters, and just 9% of the combined biomedically and contextually complicated encounters. Second, there was no guarantee that they would provide biomedically or contextually appropriate care even when they successfully probed and prompted the USP to reveal the underlying problem. For instance, even when the physician learned that a patient could not afford his or her asthma medication, there was still more than a 50% chance that the physician would prescribe a higher dosage rather than switch to a cheaper generic. Third, clinicians tended to favor biomedical information over contextual information even when both were equally important to getting the care plan right. Specifically, they probed biomedical red flags 63% of the time and contextual red flags 51% of the time. All of these differences were significant (P < .05).
One of the most unexpected findings was that those encounters in which physicians probed contextual red flags, identified contextual factors, and addressed them in the care plan were not on average any longer than those in which they did not.20 Although surprising, the likely reason becomes clear when one listens to high-performing clinicians: For instance, on hearing the comment “It’s been tough since I’ve lost my job,” the contextually sensitive clinician would promptly ask, “How has it been tough?” and, on learning about insurance problems, switch the patient to a less costly generic. This is in lieu of a misguided discussion of the need to add additional medications or increase dosages. In sum, contextualizing care is not necessarily a longer process, but it requires sensitivity and responsiveness to contextual information.
A limitation of these findings is that they are based on an analysis of physician performance across just four scenarios, each portraying one of four contextual factors. Other cases could be “harder” or “easier” depending on, for instance, the contextual factors scripted. Nevertheless, the factors selected—low health literacy, inability to afford medications, caretaker responsibility, and nutritional deprivation—are well-documented problems presenting in ambulatory care.21–24
Do Contextual Errors Impact Health Care Costs?
Inattention to context can lead to overuse and misuse of medical services. For instance, in one case a USP presented to multiple clinicians as an elderly gentleman with unexplained weight loss due to poverty and malnutrition. There were four contextual red flags pointing to a social cause of his condition. Physicians who identified the underlying problem referred the patient to social services (e.g., “Meals on Wheels”). Those who missed the clues ordered an extensive battery of tests to evaluate for malignancy.20
In a secondary analysis of the USP study, the costs of errors were computed.6 For comparison purposes, frequency and types of biomedical errors were also tabulated. Whereas the median cost of biomedical errors across all encounters with errors was $30, it was $231 for contextual errors. In sum, when physicians were challenged with clinical situations complicated by biomedical or contextual factors, errors resulting from inattention to the latter were more costly than from the former. Because these costs reflect the consequences of inappropriate care for the particular cases used in this study, neither the absolute nor relative costs should be generalized.
How Often Does Patient Context Matter to Care Planning?
In a subsequent study, we invited real patients to carry concealed audio recorders, from July 2009 to November 2012, into their encounters at two larger Veterans Affairs (VA) internal medicine ambulatory practices in the Chicago area.25 Unlike the USP study in which every “patient” presented with hints that contextual issues might underlie problems in their clinical presentations, there was no way to know in advance how often that would occur in actual practice. Hence, the team developed a coding schema, Content Coding for Contextualization of Care (“4C”), that assessed each visit by listening to the audio recording and reviewing the medical record for the presence or absence of contextual red flags.18 When a contextual red flag was present, the coders followed the same protocol discussed above, listening for physician probing, contextual factors, and, when indicated, whether care plans were contextualized.
Across 774 audio recorded encounters with 139 physicians, there were 403 encounters with contextual red flags (52%), from among which 208 contextual factors were confirmed. Care was contextualized in 59% of these, meaning that inattention to context leading to inappropriate care plans occurred in the remaining 41%.25 These findings from real cases affirmed the high incidence of contextual factors, confirming that much care planning hinges on whether care is contextualized.
Note that the frequency of contextual red flags and contextual factors reflects the particular population of patients in this study, all of whom were veterans and many quite poor. The proportion of visits in which attention to context is essential to care planning might be lower, for instance, in a more affluent, resource-rich population. In addition, the contextual error rates of the participating physicians cannot be generalized.
Is Contextualized Care Associated With Better Health Care Outcomes?
Following the index visit at which the clinician’s attention to context was measured, members of the research team not involved in coding physician performance tracked sentinel patient health care outcomes for nine months after the initial presenting red flag.25 For instance, if a patient presented with loss of control of diabetes as evidenced by a rising HbA1c, the case was scored not only on whether the physician addressed the underlying contextual factor (e.g., deteriorating vision in a patient no longer able to read his insulin syringes correctly) but also on whether diabetes control improved over the subsequent nine months of follow-up. The coders charged with scoring health care outcomes were blind as to whether the patient’s physician had been coded as contextualizing care versus making a contextual error. From among the 157 encounters for which outcomes data were available, health care outcomes improved in 71% of those encounters in which care was contextualized and in just 46% of those in which it was not—a significant difference (P < .05). These findings demonstrate that contextual errors are consequential because they predict health care outcomes.
Can Physicians Improve at Contextualizing Care?
As illustrated in the USP study, physicians prioritize biomedical over contextual information even when both are essential to care planning.20 In other words, they are more likely to attend to biomedical issues than contextual issues in patients’ lives even when both are essential to address in an effective care plan. Attempts to provide medical students and residents with a brief educational intervention to mitigate what one might term a “biomedical bias” have met with partial success. In one study, randomized fourth-year medical students received four hours of either intensive instruction or usual training during a fourth-year subinternship rotation.26 All students were then assessed using standardized patients (SPs) presenting with the same contextual red flags and contextual factors as in the USP study; SPs and assessors were blinded to student training. Those in the intervention group were more likely to contextualize care (69% versus 22%, P < .001).
Unfortunately but perhaps not surprisingly, the skills documented in the intervention group using SPs did not carry over to the actual clinical care environment in a follow-up study of residents assessed with both standardized and real patients.25 Although residents demonstrated improved performance in contextualizing care for SPs seen in a simulation center (as had the medical students), there was no improvement in performance in the clinical setting. Clearly, four hours of education is not enough to offset long-standing habits in actual practice.
Implications: Where Do We Go From Here?
What are the next steps in light of evidence of unmeasured errors that are costly, are common, and adversely impact health care outcomes? First, we recommend more widespread tracking of clinician performance at contextualizing care both to corroborate our work and extend it. It is not possible to meaningfully address a problem unless it is monitored. However, doing so will require widespread adoption of strategies for directly observing care, as detailed above.
Second, we propose using performance data on contextualization of care to drive performance improvement, through a process referred to as “audit and feedback.”27 We are piloting such a process at two VA hospitals in Illinois, where patients volunteer to audio record their care.28 Every one to two months, clinicians and their medical home care teams receive aggregate data on the proportion of contextual red flags they probed and the proportion of contextual factors they addressed. In addition, they receive a report with representative examples of effectively contextualized care and contextual errors. The strategy is to provide clinicians with the data they need to understand where they are performing well and where they are falling short at contextualizing their patients’ care. Recently, the program has expanded to include over 100 resident clinics at two internal medicine programs.
Third, we recommend efforts to avert a biomedical bias in medical school, starting in the first year through practical case-based training. Although medical schools teach about taking a social history, doing so does not necessarily build the skills and habit of thinking about patient context during the clinical encounter. Medical students and the physicians they become are too busy and too task oriented to habitually elicit information unless they appreciate how it will inform decision making. Consider if every teaching case used to illustrate biomedical principles to medical students starting in the first year of training included a contextual dimension on which appropriate care might hinge. For instance, in pharmacology, when students identify the preferred therapeutic option for a clinical presentation, they would then be told that for this patient, it was an ineffective choice. They would be challenged to determine why, prompting them to probe and assess the contextual information as outlined above until they discovered that the patient could not afford to fill the prescription. Without a contextual dimension that mirrors real life, we should not be surprised by inattention to context.
We hope a growing number of research colleagues and medical educators will join efforts to characterize and measure contextual error, devise practical ways to track clinician performance at contextualizing care, and assess strategies to intervene both through performance-in-practice improvement initiatives and preclinical curricula designed to build the skills and cognitive behaviors to provide care that is consistently sensitive to patient context. A growing body of evidence suggests that it is worth the effort and feasible.29
Acknowledgments: The authors thank Simon Auster, MD, JD, Uniformed Services University of the Health Sciences, for the formulation of the question, “What is the best next step for this patient at this time?”
1. Behforouz HL, Drain PK, Rhatigan JJ. Rethinking the social history. N Engl J Med. 2014;371:12771279.
2. National Center for Quality Assurance. Performance measurement. http://www.ncqa.org/HEDISQualityMeasurement/PerformanceMeasurement.aspx
. Accessed September 18, 2015.
3. Weiner SJ. From research evidence to context: The challenge of individualizing care. ACP J Club. 2004;141:A11A12.
4. Weiner SJ. Contextualizing medical decisions to individualize care: Lessons from the qualitative sciences. J Gen Intern Med. 2004;19:281285.
5. Kohn LT, Corrigan JM, Donaldson MS. To Err Is Human: Building a Safer Health System. 2000.Washington, DC: National Academy Press.
6. Schwartz A, Weiner SJ, Weaver F, et al. Uncharted territory: Measuring costs of diagnostic errors outside the medical record. BMJ Qual Saf. 2012;21:918924.
7. Weiner SJ. Kattan M. Contextual error. Encyclopedia of Medical Decision Making. 2009:Thousand Oaks, Calif: Sage Publications; 198202.
8. Weiner SJ. Mullner R. Medical error. Encyclopedia of Health Services Research. 2009:Vol II. Thousand Oaks, Calif: Sage Publications; 730735.
9. Engel GL. The need for a new medical model: A challenge for biomedicine. Science. 1977;196:129136.
10. Suchman AL, Markakis K, Beckman HB, Frankel R. A model of empathic communication in the medical interview. JAMA. 1997;277:678682.
11. Epstein RM. The science of patient-centered care. J Fam Pract. 2000;49:805807.
12. Bird J, Cohen-Cole SA. The three-function model of the medical interview: An educational device. Adv Psychosom Med. 1990;20:6588.
13. Stewart MA. Effective physician–patient communication and health outcomes: A review. CMAJ. 1995;152:14231433.
14. Roter DL, Stewart M, Putnam SM, Lipkin M Jr, Stiles W, Inui TS. Communication patterns of primary care physicians. JAMA. 1997;277:350356.
15. King A, Hoppe RB. “Best practice” for patient-centered communication: A narrative review. J Grad Med Educ. 2013;5:385393.
16. Kuehn BM. Patient-centered care model demands better physician–patient communication. JAMA. 2012;307:441442.
17. Weiner SJ, Schwartz A, Cyrus K, et al. Unannounced standardized patient assessment of the Roter interaction analysis system: The challenge of measuring patient-centered communication. J Gen Intern Med. 2013;2:254260.
18. Weiner SJ, Kelly B, Ashley N, et al. Content coding for contextualization of care: Evaluating physician performance at patient-centered decision making. Med Decis Making. 2014;34:97106.
19. Weiner SJ, Schwartz A, Yudkowsky R, et al. Evaluating physician performance at individualizing care: A pilot study tracking contextual errors in medical decision making. Med Decis Making. 2007;27:726734.
20. Weiner SJ, Schwartz A, Weaver F, et al. Contextual errors and failures in individualizing patient care: A multicenter study. Ann Intern Med. 2010;153:6975.
21. Schillinger D, Grumbach K, Piette J, et al. Association of health literacy with diabetes outcomes. JAMA. 2002;288:475482.
22. Nelson K, Brown ME, Lurie N. Hunger in an adult patient population. JAMA. 1998;279:12111214.
23. Arno PS, Levine C, Memmott MM. The economic value of informal caregiving. Health Aff (Millwood). 1999;18:182188.
24. Weiner S. “I can’t afford that!”: Dilemmas in the care of the uninsured and underinsured. J Gen Intern Med. 2001;16:412418.
25. Weiner SJ, Schwartz A, Sharma G, et al. Patient-centered decision making and health care outcomes: An observational study. Ann Intern Med. 2013;158:573579.
26. Schwartz A, Weiner SJ, Harris IB, Binns-Calvey A. An educational intervention for contextualizing patient care and medical students’ abilities to probe for contextual issues in simulated patients. JAMA. 2010;304:11911197.
27. Ivers N, Jamtvedt G, Flottorp S, et al. Audit and feedback: Effects on professional practice and outcomes. Cochrane Database Syst Rev. 2012;(6):CD000259.
28. Weiner SJ, Schwartz A, Sharma G, et al. Patient collected audio for performance assessment of the clinical encounter. Jt Comm J Qual Patient Saf. 2015;41:273278.
29. Weiner SJ, Schwartz A. Listening for What Matters: Avoiding Contextual Errors in Health Care. 2016.New York, NY: Oxford University Press.