Faculty Bibliography

Background: Diagnostic tests account directly for 5% of healthcare costs, but influence decisions constituting 70% of health spending. Only 5% of ordered labs are actually " high value," depending on clinical circumstances. Low-value tests, defined as not appropriate for a given clinical scenario, are ordered in one in five clinic visits. Up to $ 750 billion is spent on these low-value tests, contributing to the estimated one-quarter to one-third of healthcare spending is on wasteful services. We sought to quantify test-specific low-value ordering behaviors in urban outpatient clinics across three standardized patient cases.
Method(s): Unannounced standardized patients (USPs-highly trained actors portraying patients with standardized case presentations) were introduced into medicine residents' primary care clinics in a large urban, safety net hospital over the past five years. The USPs simulated three common outpatient clinical scenarios: a " Well" visit, a visit with a chief complaint of " Fatigue," and a visit with a diagnosis of " Asthma." Diagnostic orders were extracted via retrospective chart review for these standardized visits. For each scenario, appropriateness of diagnostic testing was determined by reference to United States Preventative Services Task Force (USPSTF) and relevant specialty society clinical practice guidelines (CPGs). " Wasteful" (over-ordered) tests were defined as those not explicitly indicated for the given scenario. Costs were derived from GoodRx.com according to local ZIP codes.
Result(s): The most commonly wasteful tests for the Asthma case were CBC (8% of 170 visits) and Chem-7 (6%), though the relative risk of over-ordering TSH was 3.8x that of other scenarios. The most commonly over-ordered tests for the Fatigue case were LFTs (14% of 148 visits) and HBV (5%), with LFTs ordered up to 15-fold more frequently than in other scenarios. The most commonly over-ordered tests for the Well case were BMP (35% of 124 visits), CBC (15%), LFTs (15%), and HBV (11%) ordered at rates up to 6.3x, 2.0x, 14.2x, and 7.4x higher than other scenarios. Finally, the average per patient excess costs were $ 8.27 (+/-$ 1.76), $ 6.79 (+/-$ 4.5), and $ 23.5 (+/-$ 9.34) for Asthma, Fatigue, and Well cases respectively.
Conclusion(s): Inappropriateness in test ordering patterns were observed through USP simulated cases. Certain tests (CBC, BMP, LFTs, and HBV) were more likely used wastefully across cases. Between cases, specific tests were ordered in an inappropriate manner (such as TSH for Asthma, LFTs for Fatigue, and BMP for Well visits). The per patient direct cost of low value testing rose above $ 20 per visit for the Well visit, though the Fatigue case exhibited the most variation. Notably, this excludes downstream (indirect) costs inestimatable from standardized encounters alone. Knowledge of wasteful utilization patterns associated with specific clinical scenarios can guide interventions targeting appropriate use of testing

Background: Considerable resources are put into training physicians to be effective providers after residency. Practicing physicians are generally assumed to be more effective and more efficient than resident physicians who are still undergoing training. We capitalize on a unique opportunity to test that hypothesis using the controlled methodology of Unannounced Standardized Patients (USPs), Standardized Patients sent into clinical environments to systematically assess provider skills in the context of a standardized clinical scenario. Due to last minute scheduling changes, a small sample of attending physicians ended up seeing USPs we had intended to send to residents. In this study, we report on comparisons between how these attending physicians performed in terms of their patient centeredness, patient activation, assessment, and communication skills in comparison to residents.
Method(s): 6 USP visits were delivered to primary care clinics in an urban safety net hospital from 2009 to 2015. Of those 700+ visits, visits were completed inadvertently with 16 attendings. We selected the 16 attendings with at least 4 years of post-graduate experience and then matched them with 2 resident visits based on hospital, time period, and USP visit type (n=32 residents). In all visits, USPs completed a behav-iorally anchored post-visit checklist that assessed patient centeredness (4 items), patient activation (2 items), visit-specific assessment (10 items), and communication skills including information gathering (4 items), relationship development (5 items) and patient education (3 items). Items were rated as not done or partially done vs. well done and summary scores were calculated as % well done. Mean scores for attendings and matched residents were compared using t-tests.
Result(s): Resident and attending scores on patient centeredness (68% vs 73%), patient activation (44% vs 38%), assessment (53% vs 51%), patient education (49% vs 52%), information gathering (71% vs 78%) and relationship development (70% vs 73%) did not significantly differ (p>.05). Nor did we see any substantial differences in variances or find any outliers.
Conclusion(s): In our matched sample of residents and attendings, there were no significant differences by training level for any of the assessed clinical skills. While we viewed the inadvertent scheduling of USP visits with attendings as an opportunity to investigate the impact of training, our study is limited by the small sample size and whether we were able to create good matches. Findings may reflect ceiling effects (our checklists are too hard) or expertise-reversal effects (experts can skip some elements of the interaction and still arrive at the correct diagnosis and treatment plan). Further research, if our mistakenly-assessed attending sample increases, could explore the influence of PGY level and of patient load as attendings carry substantially heavier patient panels and see more (and probably more complex) patients per day then residents

Background: There is an increased awareness among healthcare professionals to discuss social determinant of health (SDOH) information with patients. However, the awareness does not necessarily translate into effective response to the situation. In order to better understand the nuances in such conversations between patients and providers, we reviewed qualitative responses from Unannounced Standardized Patient (USP) portraying patients with SDOH concerns who were seen as part of a study to investigate healthcare teams' management of SDOH information.
Method(s): USPs, representing six different clinical cases, were seen by residents at an urban safety-net hospital. Each case had SDOH issues (financial and housing insecurity, social isolation), and USPs were trained to provide such information in a systematic fashion in response to provider questioning. After the encounter, USPs completed a behaviorally-anchored, standardized checklist, and also entered their impressions of the encounter in free text. The focus of this study was to evaluate these comments using a qualitative approach, focusing only on those that addressed SDOH. 258 visits occurred from 2017-present, and 209 relevant comments were analyzed.
Result(s): Three general themes emerged: residents' openness to discussion of SDOH, their understanding of how these issues related to presenting concerns, and how they responded to those concerns. Some providers did not explore SDOH prompts, e.g. " I don't think she cut me off, but she quickly moved on to her next question without further delving deeper", while others were more responsive and supportive e.g., the provider " is very open to hearing my situation, I was able to fully explain my situation clearly." Such provider behavior impacted trust and connection, e.g., " Doctor X had good communication skills, but I felt like he didn't really hear my full story" There were variations in how well providers related SDOH to medical symptoms, e.g. " he completely ignored my concerns about mold at home" [asthma case] vs. " His questions centered around possible anxiety this (housing issue) might be causing me." After acknowledgement, fewer providers provided specific information or referrals to address the problem. This lack of follow-up seemed to leave USPs feeling uncomfortable. Both empathic comments and suggestions for actions influenced their sense of activation to manage their health post-visit.
Conclusion(s): Data from the USP visits indicate that there is a range of attention to and follow up on patient presentation of SDOH needs by trainees in clinical settings. Issues of both general communication skills, awareness of connection between SDOH and health, and awareness of local resources impacted provider behavior, which then had an effect on relationship with patients. The complex issues involved in addressing SDOH highlights the diverse training needs for learners

Background: For quality care, physicians must be skilled in diagnosing and treating chronic pain. Some studies have shown gender differences in how providers manage pain. And more broadly, female providers provide more patient-centered communication which in turn has been linked to patient activation and satisfaction with care. We explore, using Unannounced Standardized Patients (USPs), whether resident physician gender is associated with the core underlying skills needed to effectively diagnose and management chronic pain: communication, patient centeredness, and patient activation.
Method(s): We designed two USP cases and sent these undercover patients into primary care clinics at two urban, safety-net clinics. The USP cases were similar: a 30-35 y.o. male, presented as a new patient to the clinic with either shoulder pain induced by heavy lifting or knee pain due to a recreational sports injury. USPs completed a post-visit checklist that assessed patient satisfaction (4 items), patient activation (3 items), and communication skills (13 items) using a behaviorally-anchored scale (not done or partly done vs. well done). Summary scores were calculated for each of the three domains. Residents provided consent for their educational data to be used for research as part of an IRB-approved medical education registry.
Result(s): A total of 135 USP visits (80 female providers, 55 male) occurred between 2012 and 2018. Female providers saw 41 shoulder pain and 39 knee pain cases while male providers saw 21 shoulder and 34 knee cases. ANOVA was used to assess differences in summary scores by provider gender (male vs female) and by case portrayed (knee vs shoulder). Skills did not differ significantly by whether knee or shoulder pain case. Gender effects were not seen for patient centeredness or for patient activation; however female providers performed significantly better at relationship development (83% vs males 72% shoulder pain; 70% vs 66% knee pain case; p<.001) and information gathering (86% vs. males 72% shoulder pain; 79% vs66% in knee case; p<.016). Male providers, however, performed slightly better in patient education and counseling (65% vs 63% for shoulder and 38% vs 33% for knee cases; p<.001).
Conclusion(s): Developing a relationship and gathering information are critical to pain management and female residents performed better than male residents in these areas. Male providers performed slightly better than women in patient education and provider gender was not associated with any differences in patient centeredness or activation. In the future, we plan to link these underlying skills to pain management decisions, documentation and ultimately to patient outcomes. We suspect that patient activation may best be measured at follow-up, something not possible with our current USP methodology. Gender differences could be viewed as striking in the context of our relatively homogeneous sample (medicine residency program) and shared clinical environment/healthcare system

Statement of Problem Or Question (One Sentence): Does actionable feedback on patient safety indicators and responses to disclosed social determinants of health (SDOH) impact clinical behavior? Objectives of Program/Intervention (No More Than Three Objectives): (1) Develop/disseminate quarterly audit-feedback reports on SDoH practice behavior, focusing on elicitation of patient information. (2) Enhance our understanding of factors related to disparities in safety/quality of care. (3) Increase rates of SDoH documentation and referral. Description of Program/Intervention, Including Organizational Context (E.G. Inpatient Vs. Outpatient, Practice or Community Characteristics): We sent Unannounced Standardized Patients (USPs) with SDoH-related needs to care teams in two urban, safety-net clinics. Data collected on practice behaviors were used for cycles of audit and feedback on the quality of electronic health record (EHR) documentation, team level information sharing, and appropriate service referral. Reports contained an evolving educational component (e.g. how to recognize, refer, and document SDoH). We disseminated reports to teams (doctors, nurses, physician's assistants, medical assistants, and staff) at routine meetings and via email. Measures of Success (Discuss Qualitative And/Or Quantitative Metrics Which Will Be Used To Evaluate Program/Intervention): Three audit feedback reports have been distributed to date. Survey data was collected at two time points, 2017 (n=77) and 2018 (n=81), to assess provider attitude changes and integration of feedback into clinical practice. Measures included change in team knowledge and attitudes towards SDoH, and response to/documentation of presented SDoH (measured via post-visit checklist and EMR). Findings To Date (It Is Not Sufficient To State Findings Will Be Discussed): Preliminary data shows no change or improvement in documentation of SDoH and limited variation between firm-level responses. (1) Only 7% of providers reported feeling strongly confident in knowing how to make referrals for social needs in 2018; no improvement since 2017. (2) Despite regular report distribution, 58% of providers reported having received no formalized feedback on responding to SDoH. 24% reported maybe or not sure. (3) 86% of 2018 survey participants self-reported having referred a patient to appropriate services when a social need was identified. Our referral data says otherwise, referrals occur for less than 30% of visits with SDoH-related needs. Key Lessons For Dissemination (What Can Others Take Away For Implementation To Their Practice Or Community?): Results suggest disconnect between team data and individual reporting: most report they refer but data suggests few do. Deeper integration of reports into team processes, attachment of feedback to curricula, and increased frequency of regular feedback may be needed for accountability. These preliminary Results help refine audit feedback methodology but research is needed to understand motivation and systems barriers to referral and documentation. Future research will look at provider attitudes toward referral processes

Background: Inter-provider variability is a major source of low-value care. The dissemination of clinical practice guidelines (CPGs) has targeted this variability, yet 44% of physicians are non-adherent to CPG. This may be due to factors including exceptionalism and incentive misalignment that present a conflict between comprehensiveness and prudence in work-up. A subset of super-utilizers are notable outliers: fewer than 0.5% of physicians account for 10% of healthcare costs. Super-utilizers order labs, request consults, order imaging, and prescribe medications at rates 30%, 140%, 14%, and 25% higher than the general population. We sought to quantify provider-specific low-value test ordering behaviors across three cases.
Method(s): Unannounced standardized patients (USPs) were trained for standardized simulation of three clinical scenarios: a "Well" visit, a chief complaint of "Fatigue," and a diagnosis of "Asthma." USPs were introduced into medicine residents' clinics in a large urban, safety-net hospital. Diagnostic orders were extracted via retrospective chart review. Scenario-specific appropriateness of diagnostic testing was determined by referencing United States Preventative Services Task Force (USPSTF) and specialty society CPGs. "Excessive" tests were those not explicitly indicated for a given scenario in either USPSTF or society CPGs (versus "indicated" tests). "Discretionary" tests were those conditionally indicated, pending patient-specific factors (such as hemoglobin A1C, pending BMI).
Result(s): One or more excessive tests were ordered in 44%, 22%, and 17% of Well (n=124), Fatigue (n=148), and Asthma (n=148) encounters respectively. Percent of orders that were excessive were 18%, 8%, and 10%, respectively. On average, 1.3 (+/-1.7) excessive orders were made. Within each case, rates of excessive ordering were positively correlated with rates of indicated and discretionary ordering, and negatively correlated with rates of omitting indicated tests. For example, in Fatigue, the correlation between excessive and indicated orders was 0.38, between excessive and discretionary orders rates was 0.59, and between excessive and omitted-indicated tests was-0.25 (all p< 0.05). A similar, statistically-significant pattern was found for the other two cases. 10 (21%) and 4 (8%) of 48 residents completing all scenarios demonstrated excessive ordering at rates atleast 1 and 2 standard deviations above the mean, respectively.
Conclusion(s): Introducing USPs representing clinical scenarios revealed marked inter-provider variability. Positive associations between rates of excessive, discretionary and indicated ordering suggest tendencies for comprehensiveness over prudence. Over one-fifth of residents completing all 3 cases were high-utilizers, and nearly one in ten were super-utilizers. Awareness of provider-level ordering tendencies can guide education and interventions supporting appropriate diagnostic use

Background: Self-efficacy to manage chronic conditions affects patients' health-related behaviors and interactions with the healthcare system and therefore influences health outcomes. Few studies have explored the complex relationships between mental health, self-efficacy, and management of chronic disease. A greater understanding of these interactions could inform successful community programming for marginalized populations such as public housing residents. Harlem Health Advocacy Partners is a community health worker (CHW) program designed to close health and social outcomes gaps in residents living in New York City Housing Authority (NYCHA). This study uses survey data collected for this initiative to explore the relationship between mental health and self-efficacy to manage chronic conditions among NYCHA residents with asthma, diabetes, and/or hypertension, and assess whether key variables such as connectedness to health care, social isolation and general health influence this relationship.
Method(s): Five NYCHA housing developments were selected for the CHW intervention with five matched developments for comparison. Four-hundred adult residents with a chronic disease were recruited. Baseline intake interviews were conducted in person. Self-efficacy for managing chronic disease was measured with a 6-question scale. Depression was assessed using PHQ9 scores, a screen for the presence and severity of depression. Difficulty with mental health was assessed with questions on how difficult mental health problems made it to do work, take care of things at home, or get along with others. Bivariate analyses were conducted to assess the relationship between mental health and self-efficacy. A hierarchical linear regression model was run with mental health and other relevant variables (selected based on availability in the dataset and theoretical significance) as independent variables and self-efficacy as the outcome variable.
Result(s): Self-rated general health predicted the greatest amount of variance in self-efficacy (15.7%, p < 0.001). Mental health also contributed significantly; difficulty with mental health contributed 4.0% (p< 0.001) and depressive symptoms contributed 1.1% (p=0.03) to the variance in self-efficacy. Other variables, including demographics, type of insurance, connectedness to a primary care provider, and social isolation, were not associated with self-efficacy. Overall, the full model explained 22.5% of the variance in self-efficacy to manage chronic conditions.
Conclusion(s): NYCHA residents with mental health difficulty or depression represent a uniquely marginalized subpopulation of public housing residents, and were found to have lower self-efficacy than other residents, which may mean decreased ability to self-manage chronic medical conditions. Future research should explore relationships among mental health, self-management, and health care outcomes with the goal of augmenting targeted CHW interventions

Background: Low-value tests, defined as inappropriate for a given clinical scenario, are ordered in one in five clinic visits. Residents tend to over-order diagnostic tests to "minimize uncertainty" of presenting cases, even though these tests are not useful according to Bayesian statistics; a pursuit deemed the "Ulysses syndrome". Simultaneously, evidence suggests residents misuse preventive tests in half of relevant clinical scenarios. We sought to quantify ordering behaviors in urban primary care clinics across three unannounced standardized cases.
Method(s): Unannounced standardized patients (USPs) were trained for standardized simulation of three clinical scenarios: a "Well" visit, a chief complaint of "Fatigue," and a diagnosis of "Asthma." USPs were introduced into medicine residents' clinics in an urban, safety-net hospital. All electronic orders were extracted via chart review. Scenario-specific appropriateness of diagnostic testing was determined by referencing United States Preventive Services Task Force (USPSTF) and society clinical practice guidelines (CPGs). "Preventive" tests (such as lipid panels or hemoglobin A1C) were derived from USPSTF guidelines whereas "Diagnostic" tests (such as pulmonary function testing for Asthma or heterophile antibodies for Fatigue) were from CPGs. "Excessive" tests were those not explicitly indicated for a given scenario in either USPSTF or CPGs (versus "indicated" tests).
Result(s): Indicated tests were ordered in 29% of Well (124 encounters), 16% of Fatigue (148 encounters), and 12% of Asthma (170 encounters) cases. One or more excessive tests were ordered in 44%, 22%, and 17% of Well, Fatigue, and Asthma encounters respectively. The distribution of indicated and excessive tests for preventive versus diagnostic purposes varied by case: In Well visits, the majority (71%) of excessive testing was in pursuit of a diagnosis, while three-quarters of indicated testing was for preventive purposes. In Fatigue and Asthma visits, the reverse patterns were true: the majority of indicated tests ordered were diagnostic (81% and 68%, respectively) while the majority of excessive tests were preventive (78% and 63%, respectively).
Conclusion(s): Introducing USPs to resident clinics revealed that, for patients presenting without a chief complaint (Well visit), residents successfully ordered less than one-third of indicated tests, and over 75% of inappropriately ordered tests pursued a diagnosis. For patients presenting with chief complaints (Fatigue and Asthma), rates of appropriate ordering were even lower (16% and 12%), and tended to overlook preventive care. In these cases, inappropriate tests tended to be ordered for preventive purposes. Awareness of resident mis-utilization of preventive and diagnostic testing in distinct clinical circumstances can guide educational efforts towards evidence-based care and resource stewardship

Background: The core Entrustable Professional Activities medical school graduates should be able to perform on day 1 of residency provides a framework for readiness for residency. Communication skills are an essential foundation for these core EPAs and yet there have been few studies that describe communication competence across the UME-GME continuum. We report on our OSCE-based assessment of communication skills from the first few weeks of medical school to the first year of medicine residency.
Method(s): Assessment of communication is consistent in our OSCE program across UME and GME. Domains include Information Gathering (5 items), Relationship Development (6 items), and Patient Education (3 items) and these are assessed via a behaviorally anchored checklist (scores=% well done) that has strong reliability and validity evidence. In this study, we report on 3 multi-station OSCEs: the Introductory Clinical Experience (ICE) OSCE that occurs within the first weeks of medical school; the high-stakes, pass/fail Comprehensive Clinical Skills Examination (CCSE) OSCE that is fielded after clerkship year; and the Medicine Residency Program's PGY 1 OSCE. Across 3 classes of medical school (2014-2016) we have complete data for the 24 students who continued on in our Medicine Residency (and who provided consent to include their educational data in an IRB-approved registry). Analyses focus on differences in communication skills over time and between cohorts and the relationship between communication skills measured in medical school and those assessed in residency.
Result(s): Communication scores show significant improvement through medical school (but not into residency) in Information gathering (ICE mean=56%; CCSE mean=76%; PGY1 mean=77%) (F=11.54, p<.001, ICE< CCSE) and in relationship development (ICE=59%; CCSE=78%; PGY1= 74% (F=10.68, p<.001, ICE < CCSE). Mean patient education skills, however, increase significantly across all 3 time points (32% to 50% to 65%; F=31.00, p<.001). Patterns are similar across cohorts except that the Class of 2016 means increase from CCSE to PGY1. Regression analyses show that CCSE information gathering scores are more strongly associated with PGY performance than ICE scores (Std Beta=.32 vs.06), while for relationship development, it is the ICE scores that are more strongly associated (Std Beta=.40 vs.24). ICE and CCSE patient education skills have associations with PGY1 skill of similar size (Std Beta=.30 and.28).
Conclusion(s): Findings, despite the small sample, suggest a clear developmental trajectory for communication skills development and that information gathering and patient education skills may be more influenced by medical school than relationship development. That communication skills seem to level out in PGY1 highlights need for re-consolidation as clinical complexity increases. Results can inform theory development on how communication skills develop and point to transitions where skills practice/feedback may be particularly important

Background: It is well documented that few ordered tests are " high value" a significant percentage of those ordered are " low-value." Residency offers an opportunity to teach high-value care and educational interventions to do so have been effective. However, the relationship between high-value care and residents' ability to communicate effectively with patients has not been explored. Ability to establish rapport, gather information effectively, and be patient-centered may impact residents' use of tests. We hypothesize that residents with poor skills in these areas may order tests less efficiently.
Method(s): Unannounced Standardized Patients (USPs) were introduced into residents' primary care clinics in a large urban, safety net hospital to portray 3 clinical scenarios: a well visit, a chief complaint of fatigue, and a diagnosis of asthma. Orders were extracted via chart review. Appropriateness of orders was determined by reference to United States Preventive Services Task Force (USPSTF) and clinical practice guidelines. Excessive tests were defined as not explicitly indicated for the scenario-indicated tests were the converse. Number of excessive and % of indicated tests were calculated across the 3 visits for 48 residents. Communication skills in information gathering (5 items) and developing a relationship (6 items) and a patient-centeredness score (4 items: took a personal interest, answered all my questions) were computed as % of behaviorally anchored items rated as " well done" and included in regression models predicting test utilization.
Result(s): On average, residents ordered 15% of indicated tests (SD 9%, 0-38%) across the 3 visits and a mean of 1.3 unnecessary tests (SD 1.7, 0-6). In the regression model, the 3 skills explained 16% of variation in unnecessary tests (p=.047). Information gathering explained the greatest share (8%, p=.041). With all 3 variables in the model, patient-centeredness was positively associated with unnecessary tests (Std Be-ta=.42, p=.016) and information gathering was negatively associated with unnecessary tests (Std Beta=-.34, p=.041). Mirroring these Results, superutilizers (10 residents ordering > =3 excessive tests) had lower information gathering and relationship development scores than other residents (66% vs 75% and 72% vs 76%) but higher patient centeredness scores (80% vs 74%)-although differences were not significant.
Conclusion(s): Our findings suggest that information gathering skills may have a small influence on residents' ordering of excessive tests. Further research with larger samples (adequate power) will help clarify the effect sizes. If our Results stand, interventions for high-value care should include information gathering skills and residency programs should continue to reinforce core communication skills training. In addition, our finding that patient centeredness was associated with ordering unnecessary tests suggests that residency programs could caution residents about conflating ordering of tests with patient-centeredness