Clinical problem solving and social determinants of health: a descriptive study using unannounced standardized patients to directly observe how resident physicians respond to social determinants of health

Introduction

Clinical reasoning and decision-making are complex, multi-step processes that require skilled evaluation of and action on information gathered during a clinical encounter. Such problem-solving processes call upon a combination of analytic skills, knowledge, and reasoning, all of which are developed through medical training and practice. A rich body of research, based on well-designed, rigorous studies focusing on fixed aspects of clinical decision-making, has documented how cognitive biases may interfere with diagnosis and treatment [1], [2]. Others, however, have called for the use of a more ecologically valid paradigm to complement these studies – one that accounts for the adaptive and inter-dependent nature of decision-making, the ways in which clinical decisions are often made under real world constraints (e.g., time, stress, and resources), the influence of the physician-patient relationship, and the norms of the clinic [3].

The need to more fully explore clinical problem-solving processes is perhaps greatest in areas where physicians’ traditional scope of practice is expanding. While psycho-social aspects of patients’ lives have always been an essential part of clinical care, recent appreciation for the influence of social determinants of health (SDoH) has led to increased pressure for physicians to more fully incorporate SDoH into clinical problem-solving. Social determinants, or the structurally determined conditions under which people are born, work, live, and grow, influence the healthiness of individuals’ living/ working environments, access to care and other important resources, and the effectiveness of clinical care [4]. Health care systems are now mobilizing to address SDoH. However, few studies have explored best practices for individual physicians to “address” SDoH. Addressing SDoH is a complex process that includes identifying SDoH, understanding their relationship to the clinical problem and the solution, and taking action to implement or facilitate solutions – all of which present cognitive challenges when compared to more narrowly-defined medical problems. Patients are not mere “vessels” of information. Physicians must use appropriate communication and interpersonal skills in order to create situations where patients are able and motivated to share critical information. Current studies, however, show that many physicians are not fully sure of how best to talk to their patients about SDoH [5], [6], [7], [8], [9], [10].

If we posit that what physicians are expected to “do” in response to SDoH is not yet well defined, then, as suggested by Croskerry [11], physicians’ cognitive dispositions to respond may have outsize influence. Cognitive disposition to respond (CDR) is a catch all, non-judgmental way of framing the factors that may determine how physicians approach clinical problems. “CDR” encourages an analytic approach to understanding how physicians reach particular sets of clinical decisions for given patients. The broad range of factors considered within CDR include the context of care, past experiences (expertise, availability heuristic, and biases), patient factors, fatigue/sleep, team characteristics, and affective state [11], [12], [13], [14]. These factors, alone and in combination, influence how physicians tackle SDoH in their clinical decision-making.

We developed a program of Unannounced Standardized Patient (USP) assessment and feedback in order to refine the training curriculum and provide residents with targeted data on their practice behavior surrounding SDoH. In particular, we introduced an asthma scenario where the patient was concerned about the role of bathroom mold in triggering worsening asthma symptoms. We chose asthma because it is a chronic condition closely linked to factors associated with structural inequalities in housing conditions (e.g., mold, poor ventilation, and air quality) [15]. We expected virtually all of our residents to uncover the mold issue. The patient was primed to discuss her mold concerns if asked about her housing situation in the context of asthma. However, we found that many residents either did not elicit the issue or if they did, they did not adequately incorporate it into their clinical decision-making.

Therefore, we sought to understand our residents’ clinical problem-solving processes. Because Unannounced Standardized Patients (USPs) provide an opportunity to directly observe physician response to a standardized clinical problem (standardized scenario, history, and patient characteristics) in their actual, real-world clinical settings, free from observation bias (unannounced), we used this methodology to describe resident physicians’ clinical problem-solving in terms of identifying and incorporating an essential social determinant into clinical decision-making.

Following Croskerry’s call to perform cognitive and affective “autopsies” (root cause analyses) in the face of diagnostic “failures,” [11] we conduct a cognitive and affective “biopsy” (the “living tissue” of clinical problem solving in real world contexts) to take a deep dive into how a sample of resident physicians respond to the clinical problem-solving challenge of addressing SDoH. While the USP methodology does not tell us everything about physicians’ clinical-problem solving (e.g., internal cognitions, mood, fatigue, knowledge, and previous experience), it provides in-depth information on what happens during physician-patient encounters (while controlling for patient presentation) and therefore serves to supplement the limited views of clinical decision-making provided by sole reliance on chart reviews and other clinical data [16], [17].

Our broad aim is to describe patterns of clinical problem-solving in the context of a social determinant directly linked to symptoms and solutions. We seek to answer three core questions: 1) Do resident physicians include SDoH in their assessment? 2) Are SDoH incorporated into diagnostic thinking and if so, do they document as part of their clinical reasoning? 3) And do physicians directly address SDoH in their “solution” (treatment plan)? We then explore the results in the context of “cognitive dispositions to respond,” identifying approaches to this particular clinical problem that suggest the need for additional education and training.

Methods

From 2017 to 2019, USPs were sent to two public, safety-net adult ambulatory practices to assess medicine resident physicians’ practices in caring for a patient presenting with a common SDoH-related issue. Data reported covers residents who provided consent as part of our IRB-approved Medical Education Research Registry.

USPs were trained to portray a 20-something female asthmatic new patient with an increasing frequency of attacks (Table 1). She works a service job, lives with her boyfriend in a decaying, moldy apartment, and fears public asthma attacks due to recent acute symptomology. At the time of the visit, the patient has attempted to contact her landlord to resolve her mold concern but was ignored. USPs shared substantial information on their chief complaint but discussed their poor quality housing only when explicitly asked, although providers need not ask directly about housing-related triggers. Evidence-based guidelines are clear that exposure to mold has long-term effects on asthma prognosis [18], [19], that timely removal of indoor mold is important to prevent symptoms [20], and that longitudinal follow-up is essential for prevention and management of symptoms [21], [22], [23], [24]. The patient does not understand differences between “rescue” and control inhalers. Minimum appropriate treatment in this case is defined as educating the patient about inhaler use, prescribing a control inhaler (Flovent) and ensuring timely follow-up. Residents may also refer to the Asthma clinic. Actors received 4 h of training to ensure consistent case portrayal and 2 h of checklist training to ensure accurate assessment.

Provider practices in eliciting, responding to, and documenting the presented housing concerns were collected via two tools: a behaviorally-anchored post-visit checklist completed by the USP and post-visit chart review of the clinical note conducted by trained staff (Table 2). USP-assessed items focused on how well the provider addressed their housing concerns during the visit. USPs were asked to elaborate on these practices via open-ended prompts for explanation. Chart review items included review of whether and how residents documented housing concerns in the history of present illness (HPI) and/or treatment plan and then actual actions taken to address the mold (e.g., referral to social services). Ongoing quality assurance monitored consistency of portrayal and accuracy of assessment using digital recordings from devices hidden in USPs belongings. The detection rate, as monitored through routine resident surveys, is low (about 10%) and almost always occurred after the visit.

Analyses focused on first describing resident physicians’ practices in response to the visit, then grouping those into broad patterns (i.e., 1) neglectors, 2) negative elicitors, 3) full elicitors), and finally, conducting in-depth analyses of those patterns in terms of resident characteristics (PGY/program), specific practice behaviors (assessment, treatment plan, and actions) and written documentation (HPI, treatment plan), and USPs’ open-ended explanatory comments. Patterns were then interpreted in light of possible biases and pre-dispositions that may have shaped residents’ approaches.

Results

There were 79 visits to categorical (CAT) and primary care (PC) medicine residents between 2017 and 2019. Eighty-six percentage of those residents provided consent (n=68). Residents were distributed across three post-graduate years (PGYs): PGY 1 16%, n=11; PGY 2 46%, n=31; and PGY 3 38%, n=26. 74% (n=50) of visits were at a hospital-based and 26% (n=18) a community-based ambulatory practice, both within the same healthcare system.

In terms of treatment decision-making, almost all residents recommended the patient use and prescribed a preventive inhaler (Flovent) (93%; 63/68); about half educated the patient about how asthma medications work (37/68), though only 20 fully assessed the patient’s understanding and corrected her misunderstanding. About half referred the patient to the asthma clinic (32/68). Of those that asked about housing, 22/54 documented mold in the patient’s note (41%).

Figure 1 summarizes how residents addressed the patient’s mold situation. A majority asked about housing (54/68), but 21% did not (14/68). In visits where the resident asked about housing, 70% (38/54) of them acknowledged and explored the patient’s response that mold might be an issue. However, 11% did not ask about housing in a way that allowed the patient to fully share housing concerns (6/54) and 19% (10/54) did not respond or responded negatively when she mentioned the mold concern. Finally, in terms of addressing SDoH, of the 38 residents who obtained full information, 21% provided resources or a referral for addressing (8/38), 26% did not provide resources/referral but did document the mold in the HPI or treatment plan, and more than half (53%; 20/38) took no action. Overall, only 8 (of 68) residents engaged in best practices for all stages of addressing SDoH: fully eliciting information, exploring and acknowledging, and addressing directly by providing/referring.

Figure 1:

Resident physicians’ responses to housing issue in asthma case (n=68 residents).

Tables 2–4 provide a summary of in-depth information on the clinical problem-solving processes of residents in each of three core groups: 1) Residents who did not ask about housing (neglectors), 2) Residents who asked about housing but with negative consequences (negative elicitors), and 3) Residents who asked about housing and acknowledged mold as a possible trigger of the patient’s exacerbation of asthma symptoms (full elicitors).

Categorical (CAT) residents were over-represented among the 14 residents who did not ask about housing – 25% compared to 12% of Primary Care (PC) residents (Fisher’s Exact=66.33, p<0.01). PGY two residents were slightly more likely to neglect housing (23 vs. 18% of PGY1 and 19% of PGY3). Rates of treatment decisions were similar between the two groups around the preventive inhaler. Review of the residents’ HPI demonstrated that neglectors tended to spend more of the HPI describing symptoms/functioning (vs. triggers) and/or health maintenance elements, identified “known” triggers like cats, smoke and colds but not the possibility of new triggers, and generally did not note that the patient had recently moved to a new apartment. Those who fully elicited the mold concerns tended to note the recent move, emphasized the change in/worsening of asthma symptoms, distinguished between “known” or “usual” triggers and new triggers, and focused on asthma rather than health maintenance and/or review of systems.

Review and content coding of USPs’ comments about neglectors identified three major plausible explanations in 11/14 visits: the visit was rushed (4/14: e.g., “Resident did say [he/she] was behind schedule and most likely rushed through”), the visit was not sufficiently patient-centered (5/14; e.g., “[she/he] seemed to be listening only enough to ask [their] question”), and the resident was side-tracked by a competing focus (2/14: e.g., the computer “staring at the computer screen for too long” and alternative trigger for asthma “really unpacked the cold with me”).

Table 3 provides USP comments for negative elicitors who asked about housing but did so in ways that had negative consequences for solving the clinical problem because they did not permit the patient to reveal mold concerns (n=6) or did not respond or responded negatively (n=10). In the former group, use of close-ended questions appeared to be the main culprit (cited by the USP in 5/6 visits). In one case, the USP reported not being able to explain concerns because the physician did not respond when she raised those concerns. Other negative elicitors (4/10) appeared to have missed the connection (“didn’t believe my housing situation was affecting my health”) – and did not document “mold” in their note. The other six negative elicitors were “non-responsive” -- the patient couldn’t tell what the physician thought about this issue. 4/6 of these residents documented mold as a possible trigger in their note.

Table 4 provides USP comments for full elicitors – those who found out about the mold and fully explored it (38/68). Full elicitors were divided into three groups – those who did not take action (n=20), meaning they did not document or provide resources/referrals, those who ONLY documented (n=10), and those who took action (n=8) by providing resources/referrals to address the mold. PC and CAT residents were equally distributed across these three groups but PGY2s were over-represented in the no action group (76% of PGY2s compared with 38% of PGY1s and 31% PGY3s). Among those who didn’t take action, three themes were apparent: 1) the sense that there was not much that could be done to address the mold (10/20; “sympathized and chalked it up to being a classic New York City housing problem”); 2) that the problem was the patient’s responsibility (4/20; “something I should try to get fixed”); and 3) the mold was simply not really responded to (2/20; “sort of glazed over” and “it didn’t seem to register”). One resident explicitly prioritized the mold as something to be addressed once the medication issues were resolved.

Among the full elicitors who documented the mold but did not provide resources, similar themes emerged. The “what can be done?” theme was present for one resident where the patient reported that failure to provide resources seemed to be a knowledge gap (“didn’t know of any resources [she/he] could point me to”). Half of (5/10) encouraged the patient to address the mold problem themselves. And in three of the visits, USP comments suggested that the resident was more focused on the diagnosis than the solution (“focus was more on the connection [with the mold]”).

Lastly, among full elicitors who took action (provided resources or made a referral) (n=8), the USP comments revealed little directly about the underlying thinking of the residents. However, comments did reflect a sense of urgency (“VERY concerned about my housing” and “mentioned that it was imperative that my super take a look”).

Discussion

We utilized a mixed-method approach to explore resident practices when caring for a patient whose SDoH is integral to diagnosis and treatment. Residents often overlooked connections between housing and the presenting illness. Our findings align with previous research on SDoH being frequently ignored during the clinical encounter [25], [26]. Among SDoH responders, few took action and documentation was low. Thematic analyses of comments based on resident groupings (neglectors, negative elicitors, full elicitors) provided further insight into clinical reasoning influences. In many cases, residents overlooked the critical diagnostic clues needed to make the appropriate connections, which, had this patient been real, would have directly affected patient care.

USP-reported experience with negative elicitors reflected aspects of common clinical biases [1], including “pre-mature closure” or “ascertainment” biases -- or a pre-decided assessment of the situation followed by interacting with the patient in ways that confirmed initial assessment, a CDR that appears to be heightened when the visit is rushed or computer use pre-occupies the resident. Such biases pose substantial risk to patient safety by increasing diagnostic error [27]. We discovered instances where neglectors intended to increase medication regimens to include steroids or additional inhalers because they did not recognize the role of the mold, perhaps because of an overly exclusive focus on only the asthma symptoms and not the environment. Educating medical trainees to be open to diagnostic clues from patients with SDoH-related concerns could prevent ineffective, unsafe treatment. Biases go beyond ascertainment and pre-mature closure and include anchoring effects -- the tendency to follow initial impressions of a patient’s presentation--among many others. Reducing diagnostic error directly associated with biases and affiliated CDR requires a multi-pronged educational approach that addresses awareness of biases and impact on patient outcomes through direct feedback and training on best practices [2]. Our results provide insight into curricular enhancements needed to train our residents to be effective clinical problem solvers in the context of SDoH.

The USP methodology provided opportunities to look closely at communication skills. Poor information gathering skills (e.g., use of close-ended questions or dismissive responses) prevented residents from collecting essential information. And USPs noted that while they suspected some “negative elicitor” residents appreciated the importance of mold (some even documenting in their note), they did not share that with the patient – which would have left a real patient to conclude that the mold probably wasn’t a real problem and/or to undermine their understanding of their disease. The role of patient-provider communication in history taking, patient understanding of disease, treatment adherence and patient outcomes is well-established and regularly assessed [28], [29]. Direct feedback for learners who fail to meet these communication standards is critical as residents using close-ended questions or inadvertently neglecting to give patients space to voice concerns will not be effective even if they have excellent clinical reasoning.

Full elicitors successfully uncovered and explored the full implications of mold but varied in what they did with that information. Some did nothing, perhaps because they did not know what to do or because they felt not much could be done, it was the patient’s problem, or was “out of scope” for them. From a cognitive perspective, residents may know about available options but not believe in their efficacy [30]. Or some residents might be predisposed toward less activist approaches to care, more generally. Our data cannot directly answer these questions but suggest avenues for future research.

Those full elicitors who did document but did not take action at least recorded their clinical reasoning, setting the stage for addressing mold in subsequent visits, although it may also be that these residents struggled to carry that reasoning over into planning “solutions.” On the other hand, we know that failing to document SDoH is widespread [25], [26] – and has led to the embedding of structured SDoH assessment into care (e.g., EHR). Our study pinpoints that failure to document is not the only problem. Residents need more training on why tracking SDoH from initial assessment forward is essential to long-term care in order to make clinical connections between specific SDoH and their implications for health. In addition, while our findings provide support for using structured screening tools [26], [31], specifically for patients who are able and willing to share relevant information, simply including SDoH fields in the EHR may not be sufficient. Some or our residents asked about housing in ways that made it hard for the patient to share specific concerns.

Ultimately a very small group took action by providing a resource or referral. Were they particularly concerned about the likely impact of the mold on symptoms or did they simply have access to information on or reinforcement regarding SDoH resources? Did their sense of urgency spur the cognitive effort needed to retrieve information to take action? One resident reported asking a colleague about resources but the colleague couldn’t provide any information -- this resident’s clinical problem solving process may have been curtailed by social norms (or shared understanding) about available options. The discovery that only eight referrals were made even after discussion of concerns suggests the need for information on appropriate responses and/or evidence on the importance of acting promptly on SDoH. Increasing options for addressing SDoH is essential – not only to help resolve the patient’s problems but also to show physicians that intervening is possible and effective – which may serve to reinforce effective clinical reasoning when it comes to SDoH.

Our study is not without limitations. While USPs are a defined tool for describing our health system, our sample size and focus on one residency program and one healthcare system reduces generalizability. USPs present to the clinic as new patients, and are sent into an already stretched public healthcare system. Microsystem factors such as clinical load (as evidenced by the USP report of some visits being rushed or with preoccupied providers) might impact provider willingness to engage in detailed discussions on SDoH, even during visits where patients are clearly affected by their living situation. We did not collect data on residents’ internal clinical reasoning and problem-solving processes.

Conclusions

Using USPs to directly observe practice behaviors in gathering information about, documenting and taking action on a standardized housing issue closely linked to symptoms is the first piece of the puzzle needed to better understand education and training to prepare physicians to address SDoH. In our study, residents missed clues about the patient’s clinical problem, did not always use effective communication skills to gather information or educate patients, and did not take action to directly address the SDoH. From an educational needs perspective, our findings suggest that there were multiple explanations for these clinical problem solving failures – deficits in clinical reasoning, beliefs about scope of practice/responsibility, neglecting causes for a focus on symptoms, overly narrow medical perspectives, expectancy/efficacy beliefs around the futility of addressing SDoH, and poor communication skills/practices. Future research should explore the influence of the clinical microsystem (e.g., screening tools, availability of resources, team workflows) on physicians’ clinical problem solving as it relates to SDoH, explore individuals’ cognitive thought processes to more specifically identify biases, and evaluate interventions designed to support accurate information gathering and clinical reasoning processes in actual practice in the context of SDoH.