https://orcid.org/0000-0003-3613-8012
Figures
Abstract
Rationale
Implicit racial bias affects many human interactions including patient-physician encounters. Its impact, however, varies between studies. We assessed the effects of physician implicit, racial bias on their management of cancer-related pain using a randomized field experiment.
Methods
We conducted an analysis of a randomized field experiment between 2012 and 2016 with 96 primary care physicians and oncologists using unannounced, Black and White standardized patients (SPs)who reported uncontrolled bone pain from metastatic lung cancer. We assessed implicit bias using a pain-adaptation of the race Implicit Association Test. We assessed clinical care by reviewing medical records and prescriptions, and we assessed communication from coded transcripts and covert audiotapes of the unannounced standardized patient office visits. We assessed effects of interactions of physicians’ implicit bias and SP race with clinical care and communication outcomes. We conducted a slopes analysis to examine the nature of significant interactions.
Results
As hypothesized, physicians with greater implicit bias provided lower quality care to Black SPs, including fewer renewals for an indicated opioid prescription and less patient-centered pain communication, but similar routine pain assessment. In contrast to our other hypotheses, physician implicit bias did not interact with SP race for prognostic communication or verbal dominance. Analysis of the slopes for the cross-over interactions showed that greater physician bias was manifested by more frequent opioid prescribing and greater discussion of pain for White SPs and slightly less frequent prescribing and pain talk for Black SPs with the opposite effect among physicians with lower implicit bias. Findings are limited by use of an unvalidated, pain-adapted IAT.
Citation: Fiscella K, Epstein RM, Griggs JJ, Marshall MM, Shields CG (2021) Is physician implicit bias associated with differences in care by patient race for metastatic cancer-related pain? PLoS ONE 16(10): e0257794. https://doi.org/10.1371/journal.pone.0257794
Editor: Juan F. Orueta, Osakidetza Basque Health Service, SPAIN
Received: November 8, 2019; Accepted: August 28, 2021; Published: October 27, 2021
Copyright: © 2021 Fiscella et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data cannot be shared publicly because the study population includes 101 individuals of varying age, sex, race, ethnicity, medical specialty and geographic location. This small sample poses a significant risk for re-identification and therefore, conventional restrictions involving the removal of identifiers such as physician name, practice name/address, phone number, fax, email, Zip Code, specific dates of standardized patient visits, dates of collection and geographic locations by themselves are not sufficient to ensure confidentiality. Thus the University of Rochester Office of Human Subject Protection will require that shared data associated with the publication be restricted to broad age categories (<45, 45-55, >55 years) and all data be stripped of race, ethnicity, specialty and study site identifier. In addition, all users will be required to sign a data use agreement that provides an additional layer of protection given the potential harm. Data are available from the University of Rochester Office for Human Subject Protection (Contact: Kelley A. O'Donoghue, AVP for Human Subject Protection, Phone: 585-273-4127; Fax: 585-273-1174; Email: Kelley_odonoghue@urmc.rochester.edu; Mail: University of Rochester, Office for Human Subject Protection, Saunders Research Building, 265 Crittenden Boulevard, Suite 1.250, CU 420628, Rochester, NY 14642-0628.
Funding: Funding: This study was funded by the NIH National Cancer Institute (R01 CA155376) to CS and RE. URL to sponsor’s website: https://www.cancer.gov/. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
There is evidence that implicit racial bias affects many human interactions including patient-physician encounters [1]. Implicit bias may undermine health care provided to minority patients [2,3]. These effects may be mediated through verbal and non-verbal communication and clinical decision-making [3]. For example, there is robust evidence that Black patients are less likely to receive opioid prescriptions for pain compared with White patients [4,5]. Implicit racial bias reduces empathetic related brain activity towards the apparent pain experienced by an individual of a different race [6–8] suggesting a potential mechanism through which implicit bias might reduce prescribing opioids for pain for Black patients.
Real world observational studies generally report that physician implicit bias adversely impacts health care for Black patients [3]. In contrast, studies of physician implicit bias involving experimental vignette studies have often shown no effect on clinical decisions [9]. Standardized patients (SPs), i.e. actors trained to portray patients (sometimes referred to as “secret shoppers”), offer a unique window into clinical interactions in that patient presentations can be standardized, demographics and patient presentations can be manipulated, and SPs can be deployed with high degrees of fidelity so that they escape detection [10]. With the exception of audit studies for new patient appointments [11], we are not aware of real world experimental studies involving Black and White SPs; although such studies have been used to assess racial bias outside of health care, for example, in hiring decisions [12,13].
We designed the current study to assess whether the observational findings from real world studies would extend to a randomized field experiment. Specifically, we aimed to examine the role of physician implicit bias in actual practice through the use of a randomized field experiment in which unannounced SPs visited consenting physicians in their offices. The SPs portrayed patients with advanced cancer; the four SP scenarios differed only by race (Black or White) and by role (typical or activated) but otherwise had identical scripts and achieved high fidelity. We previously examined typical versus activated roles because there is evidence that when patients are activated implicit biases may be attenuated [14,15]. Activated patients are more likely to make their desires known through a more assertive explanation of their problems and by asking their physicians questions. However, as previously reported, this hypothesis was not confirmed, i.e. activation did not attenuate SP race effects on opioid prescribing [16].
In this study, we examined the interaction between physicians’ implicit bias and patients’ race on the physician management of a new patient who presents with previously-diagnosed metastatic lung cancer and bone pain uncontrolled by current doses of opioid medications. Based on real world observational studies of the effects of physician implicit bias on care [3] and the presumed adverse impact of implicit bias on cancer care [17], we hypothesized that interactions between physician implicit bias and patient race, i.e. physicians with higher implicit bias seeing Black SPs would tend to provide less optimal care [18]. Specifically, we hypothesized that greater physician bias would be associated with lower likelihood of renewing opioid prescriptions when the SP was Black. We further hypothesized that physicians with greater implicit bias would use less patient-centered pain communication (frequency and depth of physician responses to and exploration of patients’ expressions of concerns about pain) [19–21] and less prognosis communication among Black SPs [22,23]. Based on prior observations [21,24–26], we hypothesized that physicians with greater implicit bias would display greater verbal dominance with Black SPs. Based on prior reports that disparities in routine communication generally are absent or small [27–29], we anticipated that implicit bias would not be associated with differences in routine pain assessment (e.g. questions about onset, character, location, radiation, etc.) by SP race.
Materials and methods
Overview
We conducted an analysis of a randomized field experiment in 2012 through 2016 on the association of physician implicit attitudes on care provided to Black and White patients. The study team trained Black and White SPs for office visits at primary care and oncology practices in New York, Indiana and Michigan. Each study site received approval in conformance to the Declaration of Helsinki guidelines, i.e. Purdue University Institutional Review Board (1009009643), the University of Rochester Research Subjects Review Board (RSRB00033086), the University of Michigan Human Research Protection Program (HUM00067842) and the McLaren Health Care Corporation Human Research Protections Program (2014–00098). Details of the study have been published [18]. Fig 1 shows participant flow through the study.
Physician participants
Study investigators approached oncologists and primary care physicians individually to identify those willing to participate. In the US, primary care clinicians often provide symptomatic treatments for patients with cancer and refer patients with cancer to oncologists for anti-cancer treatments. Physicians agreed to participate in two, unannounced SP visits, a hidden audio recording of the SP visits and to respond to surveys, including a computer administered Implicit Associations Test (IAT). Physician demographics, including age, sex, specialty and race/ethnicity, were based on physician self-report. All study participants provided written, informed consent.
Standardized patients
The research team hired and trained SPs at each site to assess physician prescribing and communication. As described above, the four SP roles created were identical except for self-identified (and apparent) race and role: 1) a Black activated SP, 2) a Black typical SP, 3) a White activated SP, and 4) a White typical SP. The investigators trained SPs to describe a history of advanced lung cancer treated with radiation and to say they recently moved from another state to be closer to an adult child. Detailed medical records were created to provide imaging, pathology, and treatment reports as well as physician notes. These were sent to physician offices prior to SPs office visits.
The SP role included spontaneous reports of uncontrolled pain despite exceeding the prescribed doses of pain medication. The research team trained some of the SPs to assume an activated patient role [21,30]. This SP role included asking more direct questions about their diagnosis and treatment including pain management, asking for clarifications and redirecting the discussion when their concerns were not addressed. SPs were blinded to the study hypotheses about race, activation and implicit bias. SP visits were audio-recorded and transcribed to assess SP-physician communication. Seventeen (17) different individuals over the three sites were trained as SPs and underwent non-invasive, office care, e.g. they were trained to defer rectal examinations, blood tests and imaging.
Randomization and blinding
The statistician for the primary study randomly assigned each physician stratified by specialty to visits by two Black or two White SPs, one activated and one not. The sequence for these office visits (activated SP followed a few months later by a typical SP or vice versa) was also randomly assigned within each physician. Coders were blinded to SP race, role and study hypotheses. Physicians were blinded to study aims including the fact that SPs would be of different races.
Implicit bias measure
After completion of all SP visits, we used a pain adaptation of the race IAT to assess physicians’ implicit racial bias once, using the IAT, which is a computer-supported measure of implicit associations, often based on images of people of different races and words denoting good or bad attributes. The premise behind the IAT is that people more quickly pair concepts they associate with each other than concepts with weaker associations. The IAT has been used extensively to assess implicit racial associations [31–33]. The IAT predicts intergroup behavior independently of measures of overt racial attitudes in both laboratory and field settings [34]. Participants matched words and Black and White facial images, both pained and non-pained expression as quickly as possible using two keys (“i” or “e”). More detailed descriptions of the IAT are available [35]. Using the algorithm of Nosek et al [35], we calculated the standardized differences between incompatible versus compatible word and picture pairings for Black and White facial images, whether pained or not. IAT was calculated as the mean of the standardized differences with a higher score indicating greater implicit bias for the participant. We used a SAS macro to assess the scoring algorithm endorsed by the Greenwald et al [36]. The overall IAT showed good internal reliability (alphafull = 0.74, alphanopain = 0.72, alphapain = 0.72). See supporting information file S1 Table. Our pain adapted version of the IAT is publically available (https://osf.io/2bpe4/).
Physician opioid pain prescribing measures
We obtained copies of study physicians’ medical records and prescriptions, determined whether an opioid was prescribed and created a dichotomous variable to indicate whether or not an opioid was prescribed.
Physician office visit communication measures
Physicians consented to covert audio recordings of SP visits but were not given any information about the demographics of the SPs, nature of the role, or the timing of the visit. Audio-recordings were professionally transcribed, stripped of any racial identifiers and coded by the research team. We used the intraclass correlation coefficient (ICC) to calculate interrater reliability of the coded communication variables using a SAS macro [37].
Patient-centered pain assessment.
Research assistants coded the transcripts using the Measure of Physician Pain Assessment (MPPA), which captures frequency and depth of physician responses to and exploration of patients’ expressions of concerns about their pain (e.g. “I understand your medication is not controlling your pain”). This measure was previously developed, piloted and validated for use in outpatient cancer consultations [38].
Routine pain assessment.
Using the same methods, we also coded for routine pain assessment (e.g. standard questions regarding pain location, intensity, character, etc.). Coders noted the presence of pain assessment and patient centered pain discussion items that we then summed for a total prognosis score. Scores on each item were truncated at four so that no one item dominated the total prognosis score.
Prognosis-related discussion.
We assessed prognosis and treatment choice discussions using the Prognostic Treatment Choices Scale (PTCC), which was developed in a pilot study [39] and recently used in a large randomized intervention trial to improve communication in patients with advanced cancer [40]. These items assess physicians’ communication of diagnostic and prognostic information and treatment options with patients with metastatic disease. Sample items are, “Physician asks if patient wants to know more about the diagnosis” and “Assessing if patients understand their diagnosis.” Coders noted the presence of prognosis items, which were then summed for a total prognosis score. Scores of each item were capped at four so that no one item dominated the prognosis total score.
Physician verbal dominance.
We examined physician verbal dominance during the visit in three ways: physician use of “I” statements, conversational dominance and physician conversational cutoffs. We assessed physician word choice using the 2015 version of the Linguistic Inquiry and Word Count (LIWC) software [41]. LIWC categorizes words in transcripts by types of speech, nouns, verbs, pronouns and other categories. We used the software to calculate the use of first person singular pronouns by the physician and to calculate total physician word count. We assessed higher frequency of physician “I” statements, which is the percentage of total words spoken by physicians [42]. We also used physician proportion of total words exchanged during the visit as a second measure of conversational dominance [43]. Last, we used rater coding of physician topic cutoffs, i.e. abrupt change of topic from what the patient was discussing [44,45].
Statistical analysis
Survey and other data collected on paper were double entered into MS Excel or QualtrixTM files, which we imported into SAS 9.4 (Cary, NC) for data processing. Data management and variable creation were done in SAS 9.4 (Cary, NC). Data analysis was conducted in STATA 15 (College Station, Texas).
We used mixed-effects linear regression and mixed-effects logistic regression to estimate the effects of race and the role of SPs and physician implicit bias on the clinical or communicational outcome. We included a random intercept in these models to account for the nesting of multiple visits within physicians [46]. We operationalized IAT as a continuous variable and included interaction terms with SP race and SP role. We examined the statistical significance of slopes and cross over interactions for Black and White SPs related to physician IAT scores. We included controls within each model for type of physician (oncologist or PCP) and research site (site = 1 if last site) and SP role. We used recommended strategies for managing missing data by using STATA missing data protocols for regression and logistic regression analyses [47]. Missingness was not associated with any physician demographic variables. Physicians who did not complete the second SP visit had missing values on IAT, but that was the only relationship. Reasons for failure to complete the second SP visit included physician relocation and declines when the office was too busy to schedule another visit. Missing data were treated as missing at random. We report estimates (β/OR and confidence intervals), ICC, log ratio test and deviance for modeling goodness of fit.
We employed multiple imputations with STATA mixed and meqrlogit procedures using the multiple imputation command [48] to address missing data and compared results with a complete case analysis. We conducted several additional sensitivity analyses. We added physician race as a covariate in the models. We substituted use of “We” for “I” statements [49]. Excluding visits in which SPs were detected did not alter our findings nor did controlling for detected visits as covariates change our findings. Finally, we calculated the IAT for pain faces and non-pain faces separately and examined the interaction of IAT with SP race.
Results
SPs completed 183 visits with 90 physicians. After prompting, physicians were able to identify SPs in 15% of office visits. Descriptive data appear in Table 1. Approximately 40% of the physician sample were female, 64% were White, and the mean age was 52 years (range 29–81). Eighty-three physicians (86%) completed the IAT, giving us 159/183 (87%) SP visits with raw IAT scores (Table 1). We addressed missing values using an imputed data set. Among physicians completing the IAT, 47% percent were oncologists, 53% were primary care physicians, roughly 40% were female, 58% were White, 15% Black, 7% Hispanic and 22% South Asian with an average age of 51 years (range 29–76). The mean length of visits was 36 minutes. Visit duration did not differ by physician IAT score, SP race, or physician specialty. The intraclass correlations (ICCs) for patient-centered pain assessment, prognosis related discussion and routine pain assessment in this sample were 0.73, 0.75, and 0.85 respectively.
Our hypotheses focused on interactions between physician IAT score and SP race. However, we first present the statistically significant, non-interaction effects shown in Table 2A and 2B.
Physicians provided more patient-centered talk to active SPs than typical SPs; physicians with high IAT scores made fewer I-statements with active SPs than typical SPs. Physicians with higher IAT scores used more “I” statements in general and provided more routine pain assessment (Table 2A). Activated SPs received more patient-centered pain talk and physician “I” statements, presumably in response to more questions (Table 2A).
We hypothesized that greater physician bias would be associated with lower likelihood of renewing opioid prescriptions. This is supported by the results (Table 2B). The interaction of race and IAT was associated with lower odds (odds ratio) of renewing opioid prescriptions. We explored the slopes of this cross-over interaction (Fig 2). For Black SPs, a one-unit increase in IAT was associated with a 47% non-statistically significant decrease in the odds of receiving an opioid prescription. Notably, for White SPs, a one-unit increase in IAT was associated with a 376% increase in the odds of opioid prescription, but this did not attain statistical significance (p = .08). Among Black and White SPs seen by physicians with lower IAT scores, we observed the opposite relationship with Black SPs receiving more opioid prescriptions than Whites. Notably, the slope for IAT and opioid prescription is significantly different between White SPs and Black SPs, OR = .11 or OR = 8.90 depending on the reference group (p = .029).
We hypothesized that patient-centered communication about pain and prognosis would be lower for Black SPs when physicians had higher IAT scores. This hypothesis was confirmed for patient centered pain communication. We examined the slopes for this cross-over interaction. We found that for Black SPs, a one-unit increase in IAT is associated with a non-statistically significant decrease in the slope pain talk. For White SPs, a one-unit increase in IAT is also non-statistically significant but trending towards improved communication for White SPs. Similar to the findings for opioid prescribing, opposite relationships were observed among physicians with lower IAT scores with these physicians providing greater patient-centered pain talk to Black than White SPs. Notably, the slope for IAT and opioid prescription is statistically significantly different between White SPs and Black SPs, β = 4.09 or β = -4.09 (p = .029) depending on the reference group.
Our hypothesis that prognosis communication would interact with physician bias was not confirmed. Our hypothesis that routine pain assessment would not be associated with race and IAT was supported by our findings. Our hypothesis that physicians with higher IAT scores seeing Black SPs would be more verbally dominant was not confirmed. We did not hypothesize any interactions for activated SPs.
We conducted several sensitivity analyses. A complete case analysis yielded similar results. When we added physician race as a covariate in the models, the results were unchanged. Substitution of the use of “We” for “I” statements showed no interaction with IAT and SP race. Excluding visits in which SPs were detected did not alter our findings nor did controlling for detected visits as covariates change our findings. We assessed the IAT scores for pain faces and non-pain faces separately and examined the interaction of IAT with SP race. S1 Table in the supporting information files shows the comparison of the coefficients and 95% CI’s of these interactions. None of the pain or non-pain coefficients were significant while the interaction of IAT and SP race were significant for opioid prescribing and patient centered pain talk (Fig 3).
Discussion
Findings from this field study involving randomized assignment of Black and White SPs to primary care and physicians and oncologists, are generally consistent with those from observational studies of practicing physicians. We confirmed hypotheses related to interactions between physicians’ implicit bias and SP race for opioid prescribing and patient-centered communication about pain. Both of these interactions were driven by physicians with higher bias favoring White SPs and disfavoring Black SPs. The opposite relationships were observed among Black and White SPs who were seen by physicians who showed lower implicit bias. In these instances, Black SP were more likely to receive opioid prescriptions and receive more patient-centered pain communication than White SPs. As predicted, we did not observe such interactions for routine pain assessment.
In previous observational studies, higher implicit bias was often associated with lower quality of care for minority patients [3], including patient ratings of patient centered care [17,50], shorter visits [17] and physician verbal dominance [26,49,51]. These types of observational studies are limited by unmeasured confounding and nuances in clinical context. In contrast, previous experimental studies of implicit bias, which involved patient vignettes, have generally yielded null findings with only two of nine studies reporting statistically significant relationships between implicit bias and clinical decision-making [9]. The reasons for these conflicting findings with real world observational data are not clear but could plausibly involve differences in activation of physician rational-cognitive process in vignettes compared with reliance on more implicit processes in the real world [52]. Moreover, physicians’ racial stereotypes regarding Black patients having higher pain tolerance might also contribute to under prescribing for pain in real world settings [53,54]. Importantly, our findings suggest that greater physician bias is manifested by not just less favorable treatment of Black SPs, but that higher implicit bias is associated with more favorable treatment of White SPs. This finding is consistent with studies documenting the effects of in-group favoritism [55] and highlights the role of implicit bias on both favoring White patients and disfavoring Black patients.
We did not hypothesize higher prescribing of opioids and more patient-centered care for pain communication for Black SPs among physicians with less implicit bias. Previous studies suggest that some Whites cope with anxiety related to inter-racial interactions by over-compensating with Blacks in an effort to show they are not biased [56–58]. Possibly, this over-compensation accounts for this unexpected finding. However, we are not aware of empirical data suggesting this occurs more often among those who score lower on the IAT. This unexpected finding requires replication.
Several hypotheses were not confirmed. Physician implicit bias did not interact with physician “I” statements, prognosis talk, physician word count, as a percentage of total word count during the visit, or quick changes in topic (cut-offs). We cannot be sure if these findings reflect differences in communication patterns between real and SPs and their physicians or whether previous observational findings reflect unmeasured confounding [26,49]. While our standardization of patient behavior allowed us to isolate the effect of patient race, this decontextualizing of race likely masked the real world complexity of interactions between physicians and actual patients of different races. Further study is necessary to examine this phenomenon with real patients.
The strengths of our study include the use of an experimental design involving unannounced, White and Black SPs making office visits to primary care physicians and oncologists. Rates of physician detection of these SPs were low, suggesting that physicians treated the SPs similarly to actual patients. To our knowledge, this is the first study to use a randomized field experiment to assess the effect of physician bias on care provided to Black and White patients in any clinical context.
There are several important limitations. We trained SPs for one specific context—a new, male patient with metastatic lung cancer and pain inadequately controlled by opioid analgesics. We randomized physicians to either two Black or two White SPs thus precluding assessing of within-physician effects. We included both primary care physicians and oncologists, which, while increasing generalizability, may have masked specialty differences that our study was underpowered to detect. We used a pain adaptation of the IAT that has not been validated in published studies. It is possible that White physicians are less familiar with Black faces expressing pain compared with White faces expressing pain. However, analyses that excluded use of the pained Black and White expression showed similar results. We cannot exclude the possibility that IAT is a marker for other unmeasured physician characteristics. Further, we did not control for multiple comparisons in order to avoid false negative findings resulting from our modest sample. Last, a meta-analysis showed that IAT effect sizes were smaller when variants of the IAT and when the outcome involved resource allocation [34]. These limitations underscore the need for replication of these findings using SP methodology.
Conclusions
We observed that physician implicit bias is associated with selected, clinically meaningful, racial differences in care among new patients with uncontrolled pain related to metastatic lung cancer. Physicians’ favorable treatment to White or Black SPs depended on the level of implicit bias. Our findings have implications for clinical care and for monitoring of racial disparities in care. Our findings provide further evidence that physician implicit bias can affect clinical care provided to Black and White patients while underscoring the need for research on mitigation strategies [59] and further exploration of the role of low implicit bias in favoring Black patients. These findings also provide further justification for systems for monitoring racial disparities in health care and for deploying best practices for mitigating them [60–64].
Supporting information
S1 Table. Difference in prescribing between black and white standardized patients at different levels of IAT.
https://doi.org/10.1371/journal.pone.0257794.s001
(DOCX)
S2 Table. Results based on full IAT, non-pain IAT and pain IAT.
https://doi.org/10.1371/journal.pone.0257794.s002
(DOCX)
Acknowledgments
We thank Kathleen Silver for her proofing, editing and formatting of the manuscript and for her assistance in submission.
References
- 1. FitzGerald C, Hurst S. Implicit bias in healthcare professionals: A systematic review. BMC Med Ethics. 2017;18:19. pmid:28249596
- 2. Hall WJ, Chapman MV, Lee KM, Merino YM, Thomas TW, Payne BK, et al. Implicit racial/ethnic bias among health care professionals and its influence on health care outcomes: A systematic review. Am J Public Health. 2015;105(12):e60–e76. pmid:26469668
- 3. Maina IW, Belton TD, Ginzberg S, Singh A, Johnson TJ. A decade of studying implicit racial/ethnic bias in healthcare providers using the implicit association test. Soc Sci Med. 2018;199:219–29. Epub 2017/05/24. pmid:28532892
- 4. Meghani SH, Byun E, Gallagher RM. Time to take stock: A meta-analysis and systematic review of analgesic treatment disparities for pain in the United States. Pain Med. 2012;13(2):150–74. Epub 2012/01/14. pmid:22239747
- 5. Friedman J, Kim D, Schneberk T, Bourgois P, Shin M, Celious A, et al. Assessment of racial/ethnic and income disparities in the prescription of opioids and other controlled medications in California. JAMA Intern Med. 2019. Epub 2019/02/12. pmid:30742196
- 6. Avenanti A, Sirigu A, Aglioti SM. Racial bias reduces empathic sensorimotor resonance with other-race pain. Curr Biol. 2010;20(11):1018–22. Epub 2010/06/12. pmid:20537539
- 7. Azevedo RT, Macaluso E, Avenanti A, Santangelo V, Cazzato V, Aglioti SM. Their pain is not our pain: Brain and autonomic correlates of empathic resonance with the pain of same and different race individuals. Hum Brain Mapp. 2013;34(12):3168–81. Epub 2012/07/19. pmid:22807311
- 8. Han S. Neurocognitive basis of racial ingroup bias in empathy. Trends Cogn Sci. 2018;22(5):400–21. Epub 2018/03/23. pmid:29563059
- 9. Dehon E, Weiss N, Jones J, Faulconer W, Hinton E, Sterling S. A systematic review of the impact of physician implicit racial bias on clinical decision making. Acad Emerg Med. 2017;24(8):895–904. Epub 2017/05/05. pmid:28472533
- 10. Siminoff LA, Rogers HL, Waller AC, Harris-Haywood S, Esptein RM, Carrio FB, et al. The advantages and challenges of unannounced standardized patient methodology to assess healthcare communication. Patient Educ Couns. 2011;82(3):318–24. Epub 2011/02/15. pmid:21316182
- 11. Sharma R, Mitra A, Stano M. Insurance, race/ethnicity, and sex in the search for a new physician. Econ Lett. 2015;137:150–3.
- 12. Quillian L, Pager D, Hexel O, Midtboen AH. Meta-analysis of field experiments shows no change in racial discrimination in hiring over time. Proc Natl Acad Sci U S A. 2017;114(41):10870–5. Epub 2017/09/14. pmid:28900012
- 13. Baldassarri D, Abascal M. Field experiments across the social sciences. Annu Rev of Sociol, Vol 43. 432017. p. 41–73.
- 14. Kalauokalani D, Franks P, Oliver JW, Meyers FJ, Kravitz RL. Can patient coaching reduce racial/ethnic disparities in cancer pain control? Secondary analysis of a randomized controlled trial. Pain Med. 2007;8(1):17–24. pmid:17244100
- 15. Alexander J, Hearld L, Mittler JN. Patient–physician role relationships and patient activation: The moderating effects of race and ethnicity. Med Care Res Rev. 2014;71(5):472–95. pmid:25027408
- 16. Shields CG, Griggs JJ, Fiscella K, Elias CM, Christ SL, Colbert J, et al. The influence of patient race and activation on pain management in advanced lung cancer: a randomized field experiment. J Gen Intern Med. 2019;34(3):435–42. Epub 2019/01/12. pmid:30632104
- 17. Penner LA, Dovidio JF, Gonzalez R, Albrecht TL, Chapman R, Foster T, et al. The effects of oncologist implicit racial bias in racially discordant oncology interactions. J Clin Oncol. 2016;34(24):2874–80. Epub 2016/06/22. pmid:27325865
- 18. Elias CM, Shields CG, Griggs JJ, Fiscella K, Christ SL, Colbert J, et al. The social and behavioral influences (SBI) study: study design and rationale for studying the effects of race and activation on cancer pain management. BMC Cancer. 2017;17(1):575. Epub 2017/08/27. pmid:28841847
- 19. Cintron A, Morrison RS. Pain and ethnicity in the United States: A systematic review. J Palliat Med. 2006;9(6):1454–73. Epub 2006/12/26. pmid:17187552
- 20. Shen MJ, Peterson EB, Costas-Muniz R, Hernandez MH, Jewell ST, Matsoukas K, et al. The effects of race and racial concordance on patient-physician communication: A systematic review of the literature. J Racial Ethn Health Disparities. 2018;5(1):117–40. Epub 2017/03/10. pmid:28275996
- 21. Gordon HS, Street RL Jr., Sharf BF, Souchek J. Racial differences in doctors’ information-giving and patients’ participation. Cancer. 2006;107(6):1313–20. Epub 2006/08/16. pmid:16909424
- 22. Ilowite MF, Cronin AM, Kang TI, Mack JW. Disparities in prognosis communication among parents of children with cancer: The impact of race and ethnicity. Cancer. 2017;123(20):3995–4003. Epub 2017/09/06. pmid:28873224
- 23. Ingersoll LT, Alexander SC, Priest J, Ladwig S, Anderson W, Fiscella K, et al. Racial/ethnic differences in prognosis communication during initial inpatient palliative care consultations among people with advanced cancer. Patient Educ Couns. 2019;102(6):1098–103. Epub 2019/01/16. pmid:30642715
- 24. Peck BM, Denney M. Disparities in the conduct of the medical encounter: The effects of physician and patient race and gender. Sage Open. 2012;2(3).
- 25. Johnson RL, Roter D, Powe NR, Cooper LA. Patient race/ethnicity and quality of patient-physician communication during medical visits. Am J Public Health. 2004;94(12):2084–90. Epub 2004/12/01. pmid:15569958
- 26. Hagiwara N, Penner LA, Gonzalez R, Eggly S, Dovidio JF, Gaertner SL, et al. Racial attitudes, physician-patient talk time ratio, and adherence in racially discordant medical interactions. Soc Sci Med. 2013;87:123–31. Epub 2013/05/02. pmid:23631787
- 27. Asch SM, Kerr EA, Keesey J, Adams JL, Setodji CM, Malik S, et al. Who is at greatest risk for receiving poor-quality health care? N Engl J Med. 2006;354(11):1147–56. pmid:16540615
- 28. Williams RL, Flocke SA, Stange KC. Race and preventive services delivery among black patients and white patients seen in primary care. Med Care. 2001;39(11):1260–7. pmid:11606879
- 29. Fiscella K, Sanders MR. Racial and ethnic disparities in the quality of health care. Annu Rev Public Health. 2016;37:375–94. Epub 2016/01/21. pmid:26789384
- 30. Hibbard JH, Stockard J, Mahoney ER, Tusler M. Development of the Patient Activation Measure (PAM): conceptualizing and measuring activation in patients and consumers. Health Serv Res. 2004;39(4 Pt 1):1005–26. Epub 2004/07/03. pmid:15230939
- 31. Dovidio JF, Penner LA, Albrecht TL, Norton WE, Gaertner SL, Shelton JN. Disparities and distrust: The implications of psychological processes for understanding racial disparities in health and health care. Soc Sci Med. 2008;67(3):478–86. Epub 2008/05/30. pmid:18508171
- 32. Greenwald AG, McGhee DE, Schwartz JL. Measuring individual differences in implicit cognition: The implicit association test. J Pers Soc Psychol. 1998;74(6):1464–80. Epub 1998/07/09. pmid:9654756
- 33. Penner LA, Dovidio JF, West TV, Gaertner SL, Albrecht TL, Dailey RK, et al. Aversive racism and medical interactions with black patients: A field study. J Exp Soc Psychology. 2010;46(2):436–40. Epub 2010/03/17. pmid:20228874
- 34. Kurdi B, Seitchik AE, Axt JR, Carroll TJ, Karapetyan A, Kaushik N, et al. Relationship between the Implicit Association Test and intergroup behavior: A meta-analysis. Am Psychol. 2019;74(5):569–86. Epub 2018/12/15. pmid:30550298
- 35. Nosek BA, Bar-Anan Y, Sriram N, Axt J, Greenwald AG. Understanding and using the brief Implicit Association Test: recommended scoring procedures. PloS one. 2014;9(12):e110938. Epub 2014/12/09. pmid:25485938
- 36. Greenwald AG, Nosek BA, Banaji MR. Understanding and using the implicit association test: I. An improved scoring algorithm. J Pers Soc Psychol. 2003;85(2):197–216. Epub 2003/08/15. pmid:12916565
- 37. Shrout PE, Fleiss JL. Intraclass correlations: Uses in assessing rater reliability. Psychol Bull. 1979;86(2):420–8. Epub 1979/03/01. pmid:18839484
- 38. Shields CG, Finley MA, Elias CM, Coker CJ, Griggs JJ, Fiscella K, et al. Pain assessment: the roles of physician certainty and curiosity. Health Commun. 2013;28(7):740–6. Epub 2013/01/30. pmid:23356451
- 39. Shields CG, Coker CJ, Poulsen SS, Doyle JM, Fiscella K, Epstein RM, et al. Patient-centered communication and prognosis discussions with cancer patients. Patient Educ Couns. 2009;77(3):437–42. Epub 2009/10/13. pmid:19819098
- 40. Epstein RM, Duberstein PR, Fenton JJ, Fiscella K, Hoerger M, Tancredi DJ, et al. Effect of a patient-centered communication intervention on oncologist-patient communication, quality of life, and health care utilization in advanced cancer: The VOICE randomized clinical trial. JAMA Oncol. 2017;3(1):92–100. Epub 2016/09/10. pmid:27612178
- 41.
Pennebaker JW, Boyd RL, Jordan K, Blackburn K. The development and psychometric properties of LIWC2015 Austin, TX: University of Texas at Austin; 2015 [cited 2021 August 24]. Available from: https://repositories.lib.utexas.edu/handle/2152/31333.
- 42. Dino A, Reysen S, Branscombe NR. Online interactions between group members who differ in status. Journal of Language and Social Psychology. 2009;28(1):85–93.
- 43. Mast MS. Dominance as expressed and inferred through speaking time: A meta-analysis. Hum Commun Res. 2002;28(3):420–50.
- 44. Marvel K, Epstein RM, Flowers K, Beckman HB. Soliciting the patient’s agenda: Have we improved? [Internet]. 1999 [cited 2021 August 24]:[283–7 pp.]. Available from: pmid:9918487
- 45. Li HZ, Krysko M, Desroches NG, Deagle G. Reconceptualizing interruptions in physician-patient interviews: Cooperative and intrusive. Commun Med. 2004;1(2):145–57. Epub 2006/07/01. pmid:16808697
- 46. Binder DA. On the variances of asymptotically normal estimators from complex surveys. Int Stat Rev. 1983;51(3):279–92.
- 47. Sidi Y, Harel O. The treatment of incomplete data: Reporting, analysis, reproducibility, and replicability. Soc Sci Med. 2018;209:169–73. Epub 2018/05/29. pmid:29807627
- 48.
StataCorp. Stata multiple-imputation reference manual: release 17 [Internet]. 2021 [cited 2021 August 24]; (Nov 6). Available from: https://www.stata.com/manuals/mi.pdf.
- 49. Hagiwara N, Slatcher RB, Eggly S, Penner LA. Physician racial bias and word use during racially discordant medical interactions. Health Commun. 2017;32(4):401–8. Epub 2016/06/17. pmid:27309596
- 50. Blair IV, Steiner JF, Fairclough DL, Hanratty R, Price DW, Hirsh HK, et al. Clinicians’ implicit ethnic/racial bias and perceptions of care among Black and Latino patients. Ann Fam Med. 2013;11(1):43–52. Epub 2013/01/16. pmid:23319505
- 51. Cooper LA, Roter DL, Carson KA, Beach MC, Sabin JA, Greenwald AG, et al. The associations of clinicians’ implicit attitudes about race with medical visit communication and patient ratings of interpersonal care. Am J Public Health. 2012;102(5):979–87. Epub 2012/03/17. pmid:22420787
- 52.
Penner LA, Dovidio JF, Manning MA, Albrecht TL, van Ryn M. Doing harm to some: Patient and provider attitudes and healthcare disparities. In: Albarracin D, Johnson BT, editors. Handbook of Attitudes, Volume 2: Applications. 2nd ed. New York: Routledge; 2019. p. 152–96.
- 53. Trawalter S, Hoffman KM, Waytz A. Racial bias in perceptions of others’ pain. PLoS One. 2012;7(11):e48546. Epub 2012/11/17. pmid:23155390
- 54. Hoffman KM, Trawalter S, Axt JR, Oliver MN. Racial bias in pain assessment and treatment recommendations, and false beliefs about biological differences between blacks and whites. Proceedings of the National Academy of Sciences. 2016;113(16):4296–301.
- 55. Balliet D, Wu J, De Dreu CK. Ingroup favoritism in cooperation: a meta-analysis. Psychol Bull. 2014;140(6):1556–81. Epub 2014/09/16. pmid:25222635
- 56. Trawalter S, Richeson JA, Shelton JN. Predicting behavior during interracial interactions: a stress and coping approach. Pers Soc Psychol Rev. 2009;13(4):243–68. Epub 2009/09/26. pmid:19778939
- 57. Taylor VJ, Yantis C, Bonam C, Hart A. What to do? Predicting coping strategies following ingroup members’ stereotypical behaviors in interracial interactions. Pers Soc Psychol Bull. 2021;47(7):1084–100. Epub 2020/09/29. pmid:32981452
- 58. Murphy MC, Richeson JA, Molden DC. Leveraging motivational mindsets to foster positive interracial interactions. Soc Personal Psychol Compass. 2011;5(2):118–31.
- 59. Sukhera J, Watling C. A framework for integrating implicit bias recognition into health professions education. Acad Med. 2018;93(1):35–40. Epub 2017/06/29. pmid:28658015
- 60. Crawford C. From the American Academy of Family Physicians: the EveryONE Project unveils implicit bias training guide. Ann Fam Med. 2020;18(2):182.
- 61. Wyatt R, Laderman M, Botwinick L, Mate K, Whittington J. Achieving health equity: a guide for health care organizations. IHI white paper Cambridge, MA: Institute for Healthcare Improvement. 2016.
- 62. Burgess DJ, Beach MC, Saha S. Mindfulness practice: A promising approach to reducing the effects of clinician implicit bias on patients. Patient Educ Couns. 2017;100(2):372–6. pmid:27665499
- 63. Joynt Maddox KE. Financial incentives and vulnerable populations: Will alternative payment models help or hurt? N Engl J Med. 2018;378(11):977–9. Epub 2018/03/15. pmid:29539282
- 64. Chin MH, Clarke AR, Nocon RS, Casey AA, Goddu AP, Keesecker NM, et al. A roadmap and best practices for organizations to reduce racial and ethnic disparities in health care. J Gen Intern Med. 2012;27(8):992–1000. Epub 2012/07/17. pmid:22798211