Abstract

Study Overview

In aim 1, we developed and administered a nationally representative survey to evaluate to what degree treatment preferences for depression or type 2 diabetes vary by prior experiences of discrimination. The survey, designed in close collaboration with community stakeholders, includes a discrete choice experiment (DCE) module to elicit treatment preferences (treatment type, provider language, trustworthiness, time to provider location, and cost) and standardized survey questions about past experience of discrimination in health care, mental health status, quality of health care, and other sociodemographic characteristics. Starting in the Methods and Results section, we describe the process for implementing the aim 1 survey (pilot testing, developing the nationally representative sample with assistance from the survey firm, Growth from Knowledge [GfK]), the design of the DCE, the methods used to analyze the survey data, and the results and conclusions from 3 separate analyses of these survey data.

In aim 2, we followed up aim 1 quantitative data with qualitative interviews with survey participants to understand how prior experiences of health care discrimination influence treatment preferences, elicitation of treatment preferences, and receipt of preferred treatment.

In aim 3, we conducted qualitative interviews with clinical stakeholders to examine the feasibility of more systematically incorporating systematic elicitation of preferences and health care discrimination experiences into shared decision-making (SDM) for treatment planning.

Background

Eliciting and attending to patient treatment preferences is key to developing a therapeutic alliance when crafting treatment plans for individuals who have depression (ie, major depressive disorder).^1,2 Attentiveness to patient preferences results in better outcomes, better adherence, and more cost-effective care for depression³ and other mental health conditions,⁴ but eliciting preferences for depression treatment among racial/ethnic minorities is complex.^5-9 These groups hold diverse cultural attitudes toward mental health care, are more likely to have experienced discrimination in past health care encounters,^8,9 and face significant disparities in access to care generally^10,11 and to psychiatric medications specifically.^12-15 These factors may directly influence treatment preferences or change how patients value treatment options in tradeoffs with factors such as provider trustworthiness or accessibility.

Racial/ethnic disparities in treatment span many health conditions¹⁶ and are of particular importance for chronic diseases,^17-19 including depression²⁰ and type 2²¹ diabetes. Researchers cite patient preferences for care as a potential mechanism for differences in treatment,²² but treatment preferences may be erroneously or incompletely measured, especially among racial/ethnic minority populations.^5,23,24 Improving methods of eliciting patient preferences is fundamental to measuring disparities and improving patient-centered care, especially in the area of chronic care. To advance these methods, in aim 1, we developed a novel, generalizable, patient-centered DCE and survey to more accurately measure patient health care preferences and to account for the underlying reasons for those preferences, particularly among racially/ethnically diverse participants with depression or type 2 diabetes.

A seminal 2003 study by Cooper and colleagues²⁵ surveyed primary care patients and found that racially/ethnically diverse patients with depression were less likely than non-Hispanic/Latino or White individuals to consider medication an acceptable treatment. This paper has been cited widely to suggest that racial/ethnic differences in treatment preferences are an important driver of disparities in antidepressant prescribing and use. Other studies also using clinical samples suggest that preferences differ by gender, as well, with men preferring medication and women preferring talk therapy.^22,26,27

Preferences are also likely to play a role in patient reluctance to engage in diabetes treatment. Black and Latino patients with diabetes are more likely than their White counterparts to worry about medication dependency and adverse effects of medication,²⁸ and Mexican Americans are more likely than other ethnic groups to perceive insulin treatment as being potentially harmful.²⁹ Preferences shape intentions and behaviors; positive attitudes toward diabetes management are strongly associated with better blood glucose levels.³⁰ Racial/ethnic differences in diabetes-related outcomes may in part be the result of these differences in preferences: Black and Mexican American patients with diabetes are 50% more likely to have poor glycemic control than non-Latino White patients.³¹ Diabetes treatment disparities partly explain the 77% higher risk of diabetes for Black individuals and 66% higher risk for Latino individuals,³² but patients in these groups have higher hemoglobin A_1c (HbA_1c) levels, rates of obesity, diabetes-associated complications, and mortality rates.³² Additionally, while racially/ethnically diverse groups have higher rates of diabetes relative to White Americans, they are less likely to receive guideline-concordant care to monitor diabetes.³³ Patients with diabetes who reported discrimination by medical staff had fewer diabetes-related primary care visits and foot examinations, reduced HbA_1c monitoring, poorer patient-provider relationships, and suboptimal care management relative to patients who did not report medical staff discrimination.^34-36

New methodologies, however, may update and add nuance to these understandings of how and why preferences for depression and diabetes treatments differ across race/ethnicity and gender. For example, prior studies have relied on clinical samples, thereby excluding individuals not receiving care—a potentially important limitation when studying treatment preferences among groups for whom disparities in access to mental health care are highly pertinent.^12,13 Further, preference measures have lacked community stakeholder input and have not accounted for real-life tradeoffs, such as cost barriers or provider reliability. Finally, there has been limited exploration of the potential drivers of preference differences, such as the effects of past discrimination in health care settings,^6,8,9,37 a factor that can influence patient decision-making and outcomes in many settings.^8,9,37-40,41

The present study used an explanatory, sequential, mixed-methods design. The first stage of the study quantitatively analyzed a nationally representative survey of White, Black, and Hispanic/Latino adults with moderate to severe depression or diabetes in the United States. In the second stage of the study, we conducted semistructured qualitative interviews among patients with depression or diabetes to better understand how treatment preferences are incorporated into clinical care and to explore the mechanisms underlying our quantitative results that demonstrate that health care discrimination alters preferences for treatment. In the third stage of the study, we conducted semistructured qualitative interviews among health care providers in an integrated safety-net health care system, exploring the provider perspective on the results of the quantitative study. We also assessed the utility and feasibility of using the DCE to elicit patient preferences and previous negative experiences during treatment planning.

This report is a complete account of all research conducted as part of this PCORI-funded study. A copy of the study protocol is included (Appendix A). All aspects of the research were approved by the Cambridge Health Alliance (CHA) IRB (CHA-IRB-1038/05/16). Text in the report draws from previously published materials and is structured such that the Patient and Stakeholder Engagement, Background, Discussion, and Conclusions sections reflect the full study, whereas the Methods and Results section is organized by specific aim. Citations for previously published articles and those under review or in preparation are available in the Related Publications section.

Patient and Stakeholder Engagement

For this project, we launched a community-academic-clinical partnership focused on people with serious mental illnesses or diabetes that adheres to the principles of community-based participatory research (CBPR). Here, and in a manuscript published in Health Affairs,⁴² we summarize the process by which this partnership was developed, identify the primary facilitators of and barriers to launching it, and make recommendations for others embarking on similar partnerships.

Academic researchers associated with a safety-net health care system were interested in studying the treatment preferences of members of racially/ethnically diverse groups who have serious mental illnesses or diabetes while involving community stakeholders in the research process. Colleagues referred the researchers to a consultant who has CBPR experience and lived experience with serious mental illness (Dr Jonathan Delman), who introduced them to a local mental health peer advocacy organization (the Transformation Center, Roxbury, Massachusetts). Despite having a lack of trust in researcher intentions, the organization decided to collaborate with the researchers because of the appropriateness of the research topic and their trust in the consultant. Together, this group applied for and received a PCORI grant to study patients' preferences for treatment of depression and diabetes and how health care discrimination may influence these preferences.

The community stakeholders included members of the peer advocacy organization (people of various races and ethnicities in recovery from mental illnesses, trauma, and addictions) as well as community-clinical liaisons (community health workers and patient experience team members) at a safety-net health care system. Academic researchers included health services researchers and clinician researchers whose aim was to improve physical and mental health in diverse communities. The CBPR consultant was included to enable rapid team development.

The researchers collaborated with a peer advocacy organization rather than with individual peer advocates to strengthen community voice, reduce power imbalances, and thus build community research capacity. The organization was selected because it had stable leadership and infrastructure to support active involvement, and it had served as a community gatekeeper in other research projects. The organization is recognized as a leader in consumer-led advocacy for people living with mental illnesses and has years of experience gaining the perspective of and developing leadership among people from multiple marginalized communities who live with mental health conditions (including Black, Hispanic/Latino, and Asian and Pacific Islander Americans). Other than joint work on the proposal with the community stakeholders, little partnership was in place at the outset. To help develop partnership, 4 CBPR principles guided stakeholders' involvement: (1) inclusion, (2) education and training, (3) clear communication, and (4) individualized attention. To achieve inclusion, the partnership included as equal members racially/ethnically diverse community stakeholder representatives living with serious mental illnesses from the peer advocacy organization, together with a diverse team of community-clinical liaisons who interacted regularly with patients. Education and training efforts included cross-training sessions intended to create a more equal partnership. In these interactive seminars, community stakeholders learned about research design and methods, academic research partners learned about community and patient perspectives, and trust-building exercises were conducted to enable all team members to share their experiences with the mental health care system. One challenge we encountered with training and engagement was turnover of research staff and community stakeholders during the project period and the need to maintain continuity with onboarding personnel. This challenge was overcome by developing clear communication through regular group meetings and one-on-one meetings between academic researchers and community stakeholders. Individualized attention was given as needed and included team meetings held at community stakeholders' sites, in-person discussions (often following team meetings), and partner-specific phone calls or email communications. The CBPR consultant coordinated education and training efforts, monitored group cohesion, and worked to maintain active community stakeholder engagement.

Between September 2016 and December 2016, the full partnership team met weekly to discuss the aims of the project and the design of the survey. The first-year project deliverables included developing and fielding a national survey of treatment preferences and discrimination experiences among members of racially/ethnically diverse groups, featuring a DCE that elicited patients' treatment preferences. By working with our community partners, many of whom are individuals with lived experiences of managing depression or diabetes and navigating the health care system, we found that their input grounded the design of the survey items. The stakeholders were completely immersed in the survey design process. The team met regularly with patient stakeholders as the survey was being developed. Initial drafts of the DCE and survey questions were presented to the patient stakeholders for feedback. After pilot tests/cognitive interviews with 12 other patients, we discussed the results of the cognitive interviews with patient stakeholders and incorporated their feedback based on those results. In addition, throughout this process, patient stakeholders provided input on the readability and relevance of the DCE attributes and survey questions as well as on the overall design of the smartphone-/tablet-based survey.

Training was a key component of the community stakeholder engagement. Dr Delman, an individual with lived experience of serious mental illness and experience in implementing CBPR, led the training for community partners. The training provided nonresearch community partners the tools and knowledge to meaningfully participate in the decision-making process throughout the lifetime of the project, including the following activities: research methods, research ethics, formulation of research questions, survey design, data analysis, and dissemination of the findings. Based on in-depth discussions with our community partners, the research team created versions of the survey. We also met separately with each community partner group (ie, the Transformation Center, CHA patient partners, and CHA volunteer health advisors) to solicit their comments and refine recruitment and retention strategies. These comments were used in smaller research team weekly meetings to integrate the responses into our data analysis and manuscript preparation. Over the course of the first-year launch of the project, our stakeholders gathered for community-academic-clinician meetings 16 times. During years 2 and 3, these meetings with stakeholders involved collaborating to refine the study topic, develop questions, design the intervention and comparators, and select outcomes and measures. Data from the survey were collected, and a complete analysis of the online survey was implemented. These findings were then shared and coanalyzed with the stakeholder/research team to inform aims 2 and 3. Qualitative interview guides were designed in close collaboration with our community stakeholders, as was the interpretation and dissemination of the results.

Methods and Results

Aim 1

Develop and administer a national survey to measure treatment preferences, how preferences vary by race/ethnicity, and how preferences are associated with experiences of prior health care discrimination.

The development and administration of the survey occurred in multiple stages. The first stage was to finalize the topics and wording of the DCE and other survey questions with intensive patient and stakeholder engagement. As described previously, this was a collaborative and immersive process with community, academic, and clinical stakeholders that resulted in a readable and relevant DCE and survey. In the following sections, we discuss moving from the design of the survey to pilot testing the survey, the formation of the sample and administration of the survey, and the methods used to analyze the survey results.

Pilot Testing and Cognitive Interviews to Refine the Survey Instrument

We conducted pilot testing of the survey separately with 12 patients (4 Black, 4 Hispanic/Latino, and 4 non-Latino White) in Spanish and English at CHA Somerville Hospital Primary Care. After the pilot test, we conducted individual cognitive interviews to assess how respondents interpreted the DCE and survey items. Half of the sample population had a diagnosis of type 2 diabetes, and half had a diagnosis of depression. Based on patient feedback from these pilot surveys/cognitive interviews and in collaboration with our patient representative and stakeholder team, we refined the survey instrument to be used in aim 1 (eg, modifying scenarios depicted in the DCE). Pilot surveys/cognitive interviews took place in 45-minute sessions (20 minutes to complete the treatment preferences survey and 25 minutes for the semistructured cognitive interview and the optimism scale). After the pilot surveys, we took the opportunity to implement stakeholder/research team cross-training and to bring community stakeholders together to discuss and modify the attributes of the DCE to combine consumer and clinical perspectives in determining salient treatment attributes. We then integrated preliminary findings from our pilot and stakeholder feedback to inform the development of a nationally representative survey. Methods and results from this nationally representative survey are detailed below.

Forming the Survey Sample and Administering the Survey

This survey was designed to obtain 2 nationally representative samples, 1 of individuals living with depression and 1 of individuals living with type 2 diabetes. It was also designed to include harder-to-reach individuals who are less likely to respond to traditional survey methods. To do this, we conducted an internet-based survey using a platform called Knowledge Panel (KP) through survey firm GfK. KP uses a 100% probability-based panel that oversamples hard-to-reach populations, such as households with cell phones only and Black and Latino respondents. The total KP sample panel from which our study was drawn contained 27 931 non-Hispanic/Latino White, 4102 Black, and 5862 Hispanic/Latino households. For our study, to reach our prespecified sample size goals based on power analyses, we targeted a reduction of the overall KP sample panel to sample sizes of n = 750 each for the separate depression and diabetes samples, with both survey samples stratified equally by race/ethnicity (about 250 respondents for each racial/ethnic group—Black, Latino, and White).

KP conducted an initial screen for our study to identify the desired sample size. This initial screen included basic demographic questions as well as screeners for depression and diabetes: The Patient Health Questionnaire-9 (PHQ-9) measures depressive symptoms, with scores ≥10 indicating moderate to severe depression,⁴³ and participants were asked to self-report their diabetes status.⁴⁴ PHQ-9 scores ≥10 were associated with 88% sensitivity and 88% specificity for identifying major depressive disorder. We expected 20.1% of respondents to screen positive for moderate to severe depression⁴⁵ and approximately 10.5% of respondents to report a diagnosis of diabetes.⁴⁶

The formation of the study sample is shown in Figure 1. A total of 8081 KP panel participants were recruited to complete the initial screen. Of those, 7879 (97.5%) completed the initial screening; 187 (2.3%) withdrew before completing the screen, and 15 refused the diabetes portion of the screening (0.2%). Among those who completed screening, 5951 (75.5%) screened negative for depression and type 2 diabetes, and 1928 (24.5%) screened positive for at least 1 of those disorders: 854 (10.8%) met criteria for depression only, 896 (11.4%) indicated type 2 diabetes only, and 178 (2.3%) screened positive for both. Among those with depression, type 2 diabetes, or both, 236 respondents (12.2%) were excluded because they did not complete the screening until after the target sample size had already been reached. Of those remaining, respondents who screened positive for both depression and diabetes were randomly assigned to join those with depression only or type 2 diabetes only and complete the depression survey (n = 77) or the diabetes survey (n = 75), respectively. This approach yielded 840 respondents who completed the depression survey and 852 who completed the diabetes survey. A total of 64 respondents who initiated the depression survey withdrew before completing it, while 65 completed the survey so quickly that their responses were considered implausible according to standardized KP rules (based on completing the survey in less than a third of the median time), and their responses were deleted. This left a final sample of 711 completers of the depression survey, which constituted an 84.6% completion rate among those who initiated the survey. For the diabetes sample, 41 respondents withdrew before completing the survey, while 10 had implausibly short completion times. This left a final sample of 801 completers, a 94.0% completion rate.⁴⁷ The final total sample consisted of 1512 respondents (503 non-Hispanic/Latino White, 505 non-Hispanic/Latino Black, and 504 Hispanic/Latino), including 81 (11%) who took the survey in Spanish.

Figure 1

CONSORT Diagram of Study Sample.

We expected a 70% final stage completion rate (the final stage completion rate is the completion rate among those in the GfK panel screened as having at least mild depression or a diabetes diagnosis) based on reports from KP of previous experience with these populations of interest. Our actual response rate was higher than expected (depression survey: 84.6%; diabetes survey: 94.0%). These relatively high completion rates likely occurred because we recruited the survey respondents from a panel of individuals who were already active participants in the larger KP panel (ie, completed other surveys before and after ours and were compensated for their participation).

To ensure a high response rate and quality survey data, we incorporated several additional strategies. We developed the survey with intensive patient and stakeholder input to obtain a relevant, comprehensible, and brief questionnaire. In addition, we took advantage of KP's extensive expertise conducting this type of survey throughout the completion of the surveys, and the KP team actively helped with data quality monitoring and assisted in the survey's design to prevent missing or fraudulent data.

Methods: Designing and Analyzing the Survey

The nationally representative survey consisted of 3 modules: (1) the DCE eliciting patient preferences; (2) questions eliciting recent mental health status, past experiences with health care discrimination, and quality ratings of received health care; and (3) questions about the sociodemographic characteristics of the respondents. In this section, we describe the development of the DCE; the other discrimination, health, health care, and sociodemographic survey measures; and the statistical methods used to assess the relationships between treatment preferences and discrimination.

Development of the DCE portion of the survey

Economic theory suggests that individuals evaluate preferences among options based on evaluations of the components (“attributes”) of each option.⁴⁸ Based on this theory, empirical methods have been developed to elicit and measure the independent contribution of each attribute to individuals' preferences while holding other attributes constant.⁴⁸ We used 1 such method to elicit preferences for depression treatment: a DCE. DCEs have examined preferences in health care generally^49,50 and in mental health care specifically.^51,52

In a DCE, respondents face repeated choices between 2 or more options, with each option defined by specific values pertaining to several attributes. Respondents are presented with a series of these discrete choices, and their responses are used to estimate the independent contribution of each attribute to the respondents' overall preferences (analysis of a set of these discrete choices enables us to assess preferences for each attribute, holding all other attributes equal).

In our DCE, choices were between hypothetical providers of depression care. Providers were defined by values pertaining to 5 attributes: (1) treatment offered (medication only, talk therapy only, both), (2) reliability (always, sometimes), (3) use of understandable language (always, sometimes), (4) travel time (15, 30, 60 minutes), and (5) out-of-pocket cost ($0, $20, $75). We determined these attributes and their specific values through a collaborative process with our community stakeholder team, which included a diverse range of patient and provider representatives. Respondents were shown sets of 2 hypothetical providers at a time, each with their own descriptors, and asked to choose their preferred provider from each set. Figure 2 provides an example. An opt-out option was not provided because previous research in a health care setting has found that opting out varies by education level and that offering an opt-out option does not alter the relative importance of attributes.⁵³

Figure 2

Example of a Choice Set in Our Discrete Choice Experiment.

Using all possible combinations of these 5 attributes and their specified values resulted in 108 hypothetical providers. In each choice set, we conducted a pairwise comparison between 2 of these 108 hypothetical providers, resulting in 5778 possible choice sets. Respondents did not have to view all such choice sets, however, to provide adequate data to detect preferences. Using specialized statistical software packages and following established methods,^54,55 we were able to reduce the number of choice sets each respondent viewed to 18. Specifically, a randomized block design was developed using the AlgDesign package⁵⁴ in R statistical software to minimize the number of choice sets each respondent saw while providing sufficient statistical power. The AlgDesign package divided the 108 provider profiles into 3 comparable blocks of 36 profiles. The profiles within each block were ordered randomly to create 18 unique profile pairs, with the stipulation that pairs differed on 2 to 4 attributes. No profile was repeated, and all 108 were used. In the survey, each respondent was randomly assigned to 1 block, meaning that they saw only 18 choice sets. The efficiency and statistical power of the final design was confirmed via simulation using Sawtooth Software,⁵⁵ specifically designed for DCEs. We used in-person pilot testing, online pilot testing, and cognitive interviews to refine the instructions and attribute wording and to determine visual presentation.

DCE attribute development

To ensure that the provider attributes and attribute values would be salient to respondents, we hosted an iterative community stakeholder–engaged research process. The process was based on 4 CBPR principles adapted for consumer-clinician collaboration⁵⁶: (1) inclusion (our team includes 3 patient advocacy organizations, including 1 governed and operated solely by people with lived experience with mental health conditions); (2) education/training (training by both community and academic partners); (3) clear communication (check-ins about mutual understanding were built into all meetings); and (4) individualized attention (calls and/or meetings with each organization and individuals). This process yielded a set of attributes and values for the DCE that used modalities and language that would be both understandable and significant to respondents with depression. Though especially relevant to the development of the DCE, stakeholder engagement was used throughout the research design, analysis, and interpretation of findings.⁴¹

The goal was a minimally time-intensive instrument; the median time to complete the 18 DCE questions was 6 minutes, 37 seconds. Each of the 1512 respondents faced 18 DCE choice sets for a total maximum number of 27 216 responses. Of that number, 26 299 responses (96.6%) were received. Because of the nature of the DCE data analysis, in which each response becomes an observation, the missing 3.4% of DCE responses did not need to be imputed but rather reduced the power of the analyses only slightly.

Discrimination measures used in the survey

The survey questions used to measure perceived health care–based discrimination were adapted from a combination of a University of California, Los Angeles (UCLA)-Westat–developed scale used in Trivedi and Ayanian⁵⁷ and Dr David Williams' Everyday Discrimination Scale.⁵⁸ This scale was modified in collaboration with feedback from our community stakeholder and patient partners, who were heavily involved in the survey design process. The first discrimination item asked if survey participants were ever treated unfairly while receiving medical care because of their race/color, ethnicity, language/accent, sexual orientation, or gender. The second item, which captured vicarious health care discrimination, asked if someone close to the respondent was unfairly treated because of their race/color, ethnicity, language/accent, sexual orientation, or gender. We used weighted conditional logit models to provide estimates for (1) depression treatment preferences by race/ethnicity and gender and (2) associations between past experiences of health care discrimination and current treatment preferences.

Other mental health, health care quality, and sociodemographic measures used in the survey

Mental health status was measured using 2 previously validated instruments: the PHQ-9 measure of depressive symptoms and the 6-item Kessler Psychological Distress Scale (K6) measure of psychological distress. Self-perceived quality of medical care is based on respondents' rating of quality of care as excellent, very good, good, fair, or poor. Other quality-of-care–related measures asked respondents to reflect on the previous 12 months and to rate (using the scale never, sometimes, usually, always) whether the medical provider seen most often for treatment (1) listened carefully, (2) showed respect, (3) spent sufficient time, and (4) was easy to understand. Sociodemographic variables included as covariates in the analysis were age (continuous), race (non-Hispanic White, non-Hispanic Black, Hispanic), gender (female or male), education (less than high school, high school, some college, and college graduate/graduate degree), marital status, and employment status.

Description of 3 analyses of survey data

We conducted 3 main analyses using the nationally representative survey/DCE data. The first aim 1 analysis compared depression and diabetes treatment preferences (see Creedon et al in the Related Publications section) by race/ethnicity and gender while accounting for access to health care, provider characteristics, and past experiences of discrimination in health care settings. A respondent's treatment preferences were determined by analysis of the results of the DCE, which identified individual preferences pertaining to each attribute (treatment offered, reliability, use of understandable language, travel time, and out-of-pocket cost), accounting for the other attributes.

In the second aim 1 analysis, we used the full sample of the nationally representative survey (N = 1512), conducting a latent class analysis (LCA) using Mplus software to illustrate the intersections of personally experienced and vicarious discrimination attributes and to assess the associations of membership in latent classes of experience with discrimination with provider characteristics (eg, quality of care, effective communication, respect in the health care encounter) and mental health status (see Adams et al in the Related Publications section).

In the third aim 1 analysis conducted among only the diabetes sample (n = 801; see Flores et al in the Related Publications section), we conducted a mediation analysis to explore how provider communication style mediates the relationship between perceived health care–based discrimination and perceived care quality. The dependent variable of interest is self-perceived quality of medical care, and the primary independent variable is self-perceived discrimination, as described previously. The mediator, developed using structural equation modeling, is a latent variable labeled therapeutic relationship, composed of 4 endogenous variables that originated from responses to 4 survey questions. Respondents were asked to reflect on the previous 12 months and rate (never, sometimes, usually, always) the medical provider seen most often for their diabetic care on the frequency with which that provider (1) listened carefully, (2) showed respect, (3) spent sufficient time, and (4) was easy to understand. We used structural equation models to conduct a confirmatory factor analysis and to estimate the direct effects of the independent variable and the indirect effects mediated through the pathway of the therapeutic relationship latent variable.

Results

Below, we describe the results from the analyses that were conducted using the nationally representative survey data. We first describe the characteristics of the depression and diabetes samples. Next, we provide results from the 3 analyses of the survey. Taken together, these analyses assess associations among treatment preferences, discrimination, and race/ethnicity and view how these factors intersect to influence mental well-being and health care quality.

Sample characteristics of individuals with depression

Table 1a describes the survey sample for patients with depression. Depression severity as measured by PHQ-9 scores did not vary significantly by racial/ethnic group or gender (noting that sample inclusion required a PHQ-9 score ≥10). Compared with the non-Hispanic/Latino White respondents, however, the non-Hispanic/Latino Black and Hispanic/Latino respondents were, on average, about 4 years younger, were less likely to be a high school graduate or to be married, and had less income.

Table 1a

Characteristics of Survey Sample: Depression.

Sample characteristics of individuals with diabetes

Table 1b describes the survey sample for patients with type 2 diabetes. Compared with patients who did not report experiencing health care discrimination, those who reported discrimination were more likely to report lower diabetes care quality, to identify as non-Latino Black or Hispanic/Latino, to have earned a Bachelor's degree or higher, to be in the $30 000 to $39 999 income bracket, to score higher on the K6, and to live in metropolitan areas. Inversely, those who did not report discrimination were more likely to be in the ≥$75 000 income bracket and to live in nonmetropolitan areas.

Table 1b

Characteristics of Survey Sample: Diabetes.

Aim 1, analysis 1 results

Depression treatment preferences results

Figure 3 presents the isolated medication-vs–talk therapy preference results (derived from the DCE) from the conditional logit models comparing preferences by race/ethnicity and gender in the depression sample. On average, non-Hispanic/Latino White respondents (odds ratio [OR], 0.80; 95% CI, 0.64-0.99) and men (OR, 0.76; 95% CI, 0.60-0.96) preferred medication over talk therapy, while non-Hispanic/Latino Black respondents, Hispanic/Latino respondents, and women did not have a significant preference. Respondents preferred providers who offered both medication and talk therapy over those offering only 1 or the other. The preference differences between the non-Hispanic/Latino White respondents (who preferred medication) and the non-Hispanic/Latino Black and Hispanic/Latino respondents (who did not have a preference) were marginally statistically significant, as was the difference between men (who preferred medication) and women (who did not have a preference).⁴¹

Figure 3

Preference for Providers Offering Medication vs Talk Therapy as Treatment for Depression, by Racial/Ethnic Group and Gender.

Past health care discrimination among depression sample group

Table 2 summarizes past experiences of health care discrimination in the depression sample. Compared with non-Hispanic/Latino White respondents, the non-Hispanic/Latino Black respondents were 3 to 4 times as likely to report having experienced discrimination in health care settings, and the Hispanic/Latino respondents were 2 to 3 times as likely to report it, depending on the source of discrimination. Men and women were equally likely to report discrimination.

Table 2

Experiences of Discrimination in Health Care Settings Among People With Depression, by Racial/Ethnic Group and Gender.

Figure 4 presents, within each racial/ethnic and gender group, the isolated medication-vs–talk therapy preference results from the conditional logit models comparing preferences by whether a respondent had experienced discrimination from a medical provider, front-desk staff, or both. Full model results are provided in Appendix B: Supplementary Tables. Non-Hispanic/Latino Black respondents and women who had experienced discrimination from front-desk staff or from both medical providers and front-desk staff had significantly lower preferences for talk therapy and greater preferences for medication. The remainder of the associations between past discrimination and treatment preferences were not significant. For all racial/ethnic groups and women (though not for men), however, we observed a pattern of preferring medication over talk therapy in the presence of any kind of discrimination.⁴¹

Figure 4

Preference for Providers Offering Medication vs Talk Therapy as Treatment for Depression, by Racial/Ethnic Group and Gender and Stratified by Whether One Experienced Past Discrimination in Health Care Settings.

Type 2 diabetes treatment preferences results

White respondents preferred medication only (P < .001), but the Black and Latino respondents did not have a preference for medication vs behavior modification only. Additionally, across all racial/ethnic categories, men preferred medication only (P < .001), as did women (P = .014). When comparing medication only with a combination of medication and behavioral modification, all racial/ethnic groups and genders preferred both to approximately the same degree.

Effects of discrimination on preferences for medication vs behavior modification among patients with diabetes

Facing discrimination from a medical provider and staff in addition to having someone close to you experience discrimination shifted preferences toward behavior modification and away from medications for White and Black respondents. Among Latino respondents, the opposite was true, but this result was not statistically significant. Facing discrimination from a medical provider or staff in addition to having someone close to you experience discrimination shifted preferences toward behavior modification and away from medications for men but not for women (see Creedon et al in the Related Publications section).

Aim 1, analysis 2 results

The second analysis uses LCA to identify classes of different types of experiences with discrimination. It assesses the association between these classes of discrimination with provider characteristics (eg, quality of care, effective communication, and respect in the health care encounter) and mental health status. The aim 1, analysis 2 LCA characterized 3 classes in the full diabetes and depression sample (N = 1512): The largest latent class (85.5% of the total sample) represents low discrimination (LD) and is composed of respondents who had low item-class probabilities (relatively low endorsement of discrimination classes) of reporting personal or vicarious discrimination across all identity attributes. The second-largest class (10.1%) is characterized as moderate personal/vicarious intersectional (MPVI) and is composed of respondents with moderate item-class probabilities (moderate endorsement of discrimination classes) on personally observed discrimination and relatively higher item-class probabilities (high endorsement of discrimination classes) of reporting vicarious discrimination. Finally, the smallest class (4.4%) is characterized as high personal/vicarious race-only (HPVR) and is composed of respondents with high item-class probabilities of personally and vicariously experienced discrimination on the race attribute. When looking at the sociodemographic characteristics and mental health outcomes across the 3 latent classes, we found that the LD class had significantly older respondents (mean age = 54.6 years) compared with both the HPVR and MPVI classes (mean age = 49.7 and 44.5 years, respectively). The distribution of gender was relatively similar across classes, with a higher proportion of women in each class than men. The HPVR class had the highest proportion of non-Hispanic Black respondents (77%) compared with the other 2 classes, whereas the MPVI class had the highest proportion of Hispanic respondents (49%). Individuals in the MPVI class were also more likely to have more education (eg, some college, Bachelor's degree, or a graduate degree) and to be employed. The HPVR and MPVI classes included a significantly higher proportion of depressed cohort members than diabetic cohort members within class and in comparison with the LD class. Finally, respondents in the MPVI class had the highest average depression (PHQ-9) and psychological distress (K6) scores (mean values, 12.2 and 9.96, respectively). It is important to note, however, that the average PHQ-9 and K6 scores among respondents in the MPVI class, while the highest of the 3 classes, did not differ significantly from the average PHQ-9 and K6 scores in the HPVR class.

We assessed whether latent class membership was significantly associated with 2 mental health outcomes—depressive symptoms and psychological distress—for the pooled sample and stratified by chronic disease category, after adjusting for sociodemographic covariates. In adjusted linear regression models among individuals from the depression cohort, individuals in the HPVR and MPVI classes reported significantly (P < .05) higher depressive symptoms (β = 1.55; 95% CI, .02-3.07 and β = 1.04; 95% CI, .66-2.02, respectively) than the LD class. The MPVI class reported significantly (P < .05) higher psychological distress (β = 1.47; 95% CI, .32-2.63) than the remaining 2 classes. Similarly, among respondents with diabetes, individuals in the HPVR and MPVI classes had significantly (P < .05) higher depressive symptoms (β = 1.76; 95% CI, .13-3.39 and β = 2.10; 95% CI, .87-3.33, respectively) and psychological distress (β = 1.79; 95% CI, .16-3.44 and β = 1.82; 95% CI, .57-3.06, respectively) than the referent LD class (see Adams et al in the Related Publications section).

Aim 1, analysis 3 results

In analysis 3, conducted on the sample of individuals living with diabetes (n = 801), therapeutic relationship was confirmed as a unidimensional latent construct with excellent model fit (root mean square error of approximation [RMSEA] = 0.033; comparative fit index [CFI] = 0.997; Tucker-Lewis index [TLI] = 0.993). Adjusted results suggest that respondents reporting health care discrimination by race/ethnicity or language were more likely to report a diminished therapeutic relationship (β = −.17, P < .001), which in turn was associated with lower diabetes care quality ratings (β = .73, P < .001). Therapeutic relationships fully mediated the association with discrimination on diabetes care quality (β = −.02; bootstrap 95% CI, −.11 to .06). Our mediated path model also demonstrated excellent fit (RMSEA = 0.026; CFI = 0.978; TLI = 0.971) (see Flores et al in the Related Publications section).

Aim 2

Interview survey participants to understand how prior health care discrimination influences treatment preferences and receipt of preferred treatment.

Theoretical Framework and Interview Guide Design

The research group for this PCORI study (community-academic-clinical partnership) used a social-ecological framework⁵⁹ to inform a conceptual model of adaptive patient treatment preferences to inform qualitative interview guides and understand the types of discrimination that participants faced; the effect of these experiences; and the relationship between past experiences of discrimination and patient preferences, including how past discrimination experiences may alter the elicitation of treatment preferences (Appendix C). In this adaptive model of patient preferences, beliefs and preferences about treatments are updated based on exposure to individual-level, family-level, and sociocultural factors (Figure 5), including health care discrimination, as well as through seeking and obtaining treatment. This portion of the overall study focused primarily on the role that individual-level experiences of health care discrimination (those most proximal to the individual) have in seeking and receiving mental health treatment, shaping preferences for treatment of depression and diabetes, and the degree to which those preferences are routinely elicited and incorporated into treatment plans. Both the quantitative and qualitative data collection included questions about awareness of discrimination against family members or friends as well as whether participants believed that racially/ethnically diverse patients are always treated equally in health care (eg, awareness of more distal discrimination experiences).⁷¹

Figure 5

Conceptual Model of Adaptive Preferences Informed by Social-Ecological Model.

Qualitative Data Collection

Members of the research team were invited to join a qualitative data-collection and analysis subgroup. Six team members of the analysis subgroup conducted semistructured interviews (focused on understanding how past experiences of discrimination influence treatment preferences and preference elicitation) with 21 individuals with depressive symptoms and 19 individuals with diabetes. A PhD-level senior qualitative researcher and a PhD-level senior mixed-methods health services researcher led the subgroup, which also included 2 research assistants/coordinators, 1 student research volunteer, and 1 volunteer community health worker. Demographically, all 6 subgroup researchers were women, 5 were from racially/ethnically diverse groups, and 2 were native speakers of Spanish. This qualitative subgroup worked in close collaboration with the larger community-academic-clinical partnership on the codevelopment of interview guides, participant recruitment, data collection, and data analyses and interpretation. The qualitative subgroup met iteratively to train its subgroup members on the research protocol and qualitative data-collection methods and to practice using and refining the interview guide. Less-experienced interviewers observed 1 to 2 initial interviews conducted by the senior researchers before conducting their first interview.⁷¹

Sample

Table 3 describes the interview sample. Forty participants completed semistructured interviews, including by phone (n = 12 for depression and n = 12 for diabetes for individuals who had previously taken the nationally representative survey) and in person (n = 9 for depression and n = 7 for diabetes). Participants who completed the in-person interviews were not part of the nationally representative survey, but they were recruited locally via the community partner organization to supplement the sample size because of an unanticipated change in recruiting and scheduling costs through the third-party online survey organization.

Table 3

Characteristics of Depression and Type 2 Diabetes Interview Sample.

Two recruitment methods were used for interview participants. In the first recruitment method, the 12 individuals in each group who had previously participated in the nationally representative survey described in aim 1 were purposively sampled to achieve an even mix (4 each) of non-Hispanic/Latino White, non-Hispanic/Latino Black, and Hispanic/Latino participants, 50% of whom had reported health care–based discrimination on the survey and 50% of whom had not. People who had not experienced discrimination were included in order to understand how their depression treatment experiences differed from those who had experienced discrimination. Spanish-speaking participants were interviewed by native speakers of Spanish. Inclusion criteria included being ≥18 years of age and scoring ≥10 on the PHQ-9 (representing moderate to severe depressive symptoms).⁴³ Survey participants were invited to join the qualitative study on a first-come, first-served basis (see Appendix B for breakdowns within each category). Interviewers from the research team then phoned participants at times prescheduled by the online survey organization.

In the second recruitment method, the additional local participants were recruited through the mental health recovery advocacy partner organization via email and word of mouth. The research assistant provided additional study information to those who were contacted, issued an eligibility screener, and then scheduled the interview. Individuals were eligible if they (1) had ever been told by a doctor or nurse that they had depression or diabetes; (2) were ≥18 years of age; and (3) identified as Black or African American, Hispanic/Latino, or non-Hispanic/Latino White.

All participants provided written informed consent to be interviewed and received a $50 gift card for participation. The participants who were recruited via the second recruitment method and interviewed in person completed a brief demographic questionnaire as well as the structured survey referenced in aim 1 (a paper version identical to the survey that participants had completed online in aim 1) immediately before commencing the interview. Interviews were approximately 45 minutes and were conducted from November 6-17, 2017 (phone participants) and from January 15-19, 2018 (in-person participants).⁷¹

Data Analysis

Transcribed interviews were analyzed using a thematic analysis approach, and both deductive analysis (using codes developed in advance from interview guides) and inductive analysis (using open coding) were conducted using Dedoose.⁶⁰ The senior qualitative researcher designed an initial qualitative code tree using the codeveloped interview guide, which was refined during iterative group meetings. Six members of the study team coded the interviews (2 researchers per interview) and generated summary analytic memos, then met to discuss coding for individual interviews, review and resolve coding discrepancies, and consolidate codes for each interview (discussing these discrepancies within the larger qualitative subgroup as needed).

The qualitative research subgroup attended to issues of reflexivity (examination of and minimizing researchers' own interpretive biases). This was done within the subgroup through debrief sessions; via memos after conducting and coding each interview; and through continuous, iterative engagement with the larger research-practice-community partnership, including intentional reflection on the research process and collaborating to interpret qualitative data and prepare the manuscript.

Results

Demographic summaries of each group are shown in Table 3. Thematic results of qualitative interviews are described and discussed below. These are described separately for the depression and type 2 diabetes samples.

Types of discrimination experienced

During qualitative interviews, several additional types or examples of discrimination were shared in addition to the types of discrimination that participants were explicitly asked about in the aim 1 quantitative survey. In addition to discrimination by race/color, ethnicity, language, gender, or sexual orientation (those attributes asked about in the survey described in aim 1), participants shared stories that described discrimination by income or social class; discrimination based mental health diagnosis; and discrimination based on weight, style of dress, age, and disability status. The 2 attributes most often cited after those attributes that were captured by the survey described in aim 1 were income/social class and mental health diagnosis. We provide additional details about the type of discrimination experienced where relevant in the discussion of themes below.

Depression sample thematic results

Results from qualitative coding and thematic analysis of interviews for the depression sample are organized by major themes within each of the 3 overarching study objectives outlined in Table 4.

Table 4

Research Objectives for Aim 2 Qualitative Analysis and Major Qualitative Themes for Depression Sample.

Objective 1. Understand How Past Experiences of Health Care Discrimination Play a Role in the Experience of Seeking and Receiving Mental Health Treatment

Theme 1a: Challenges to receiving mental health treatment

Participants shared the many pathways they undertook in search of mental health treatment. Regardless of experiencing health care discrimination, it was extremely difficult to find a mental health provider who accepted the patient's insurance and was accessible to the patient. Participants also struggled to find a provider who was a “good fit” and to stay in mental health treatment settings with strict attendance policies: “Every time I called, I was told that there's a wait list, wait list, for several months, several months. At one point, they told me they are not even taking people.” (Female, White, participant 21, prior discrimination: lack of full consent to expensive treatment)

Participants often initiated mental health treatment through a primary care doctor. Here, participants' experiences demonstrated how a trusting, long-term relationship with a provider was often critical to opening up discussions of mental health treatment and in continuing to try different treatment modalities over a longer period of time: “With the talking, it was hard to find somebody that I felt I could trust. That took a lot.” (Female, White, participant 10, prior discrimination: income/social class)

Theme 1b: Health care discrimination exacerbates already complex mental health care–seeking

Health care discrimination related to race/ethnicity

When participants could recall personal experiences of discrimination and were also aware of discrimination occurring at multiple social-ecological levels, this awareness often served to amplify their personal experiences and reinforce the belief that as individuals in racially/ethnically diverse groups, their lives were not valued equally or that they were at risk of maltreatment in medical settings and thus they could not trust health systems and medical providers. This feeling led to participants expressing feelings from extreme disappointment and frustration to, at times, fear and terror of medical settings. Below we highlight 2 powerful narratives from Black women.

He left me without [sleep] meds and … I was the only Black person in the waiting room … I almost had a nervous breakdown and he's giving [these White patients] over the amount of meds they need? … Back in the 60s and 70s when Black folks were shooting dope left and right, working the street, nobody was saying nothing … America is involved in the opioid crisis because White folks is dropping dead. It's just ridiculous, it's just plain ridiculous. (Female, Black, participant 18, prior discrimination)

So I just feel like, you know, Black people, we are just like cattle. You know if they want to kill us, they kill us. If they want our body parts they'll take them and stuff like that … there's things that's happening with me right now and I'm—I'm not even going to the hospital because they killed my—my sister's man. (Female, Black, participant 6, prior discrimination)

Similarly, a Hispanic/Latino participant who reported past experiences of health care discrimination highlighted how his choice of clothing led his providers to perceive him as threatening:

They would perceive you as you're a thug, you're going to rob them, you're going to, you're going to … because you know you're—you're dressed that way. (Male, Hispanic/Latino, participant 7, prior discrimination)

Health care discrimination related to mental health diagnosis

Our research team did not include discrimination related to mental health diagnosis in the original definition of discrimination that was operationalized for aim 1. However, inductive analysis with community partners identified this as a legitimate and important domain of discrimination to include in our analyses. Participants expressed concern that, because of their mental health diagnoses, they often lost credibility to advocate for themselves. For example, patients felt that providers and staff often judged patients with a mental health diagnosis as having limited competence. In those instances, they saw the role of the physician in a trusting relationship as particularly important:

No one is going to listen because I'm mentally ill, but if my doctor says, “she is fine, she can do this,” they will all back away. They have no choice. So it's kind of like a necessity of life… . So like, basically, you know, they do discriminate. (Female, White, participant 19, prior discrimination: mental health diagnosis)

Like, oh, there was one time when my therapist asked if—it had to do with cooking—and my therapist asked if I could boil water. And I felt like saying, are you like, you know what I mean, when you just want to like storm out of the office, no, because I can't do that because they'll just up my medication. (Female, White, participant 13, prior discrimination: mental health diagnosis)

Participants experiencing distress (including in response to health care discrimination) reported that their emotional reactions were often attributed to their mental health condition and that displaying these emotions put them at risk for increased medication dosages:

I went to [a medical clinic] and I went for a sore throat, but I was told that I had another issue … and I got concerned. And she said that I seemed a little [more] worried than usual and that my doctor needs to adjust my medicine. And she just saw that I was on a psychiatric medicine and I said … “It's not for worrying, it doesn't need to be adjusted.” And I tried to put in a complaint against them.” (Female, White, participant 15, prior discrimination: mental health diagnosis)

Health care discrimination because of perceived social class

Participants often reported discrimination in health care settings related to their ability to pay for treatment, which repeatedly occurred as early as entry into the waiting room and was related to their receiving public health insurance:

If they're going to look down on me because they think I don't have that kind of money, you know, at that time, you know—if you're not a moneyed person then you don't belong in that clinic. (Female, White, participant 10, prior discrimination: income/social class)

Race/ethnicity or mental health diagnosis often intersected, however, with health care discrimination based on perceived social class. One participant described this as being a double disadvantage (“2 strikes against you”) right at the front desk and that it amplified her distrust of the clinical settings she visited for care.

But like if it's a White nurse up there doing your registration and I felt they look at you really funny especially when you pull out your [insurance] card and stuff like that you know … no telling what they might do…. You're going to get a hard time because you have [Medicaid] or Medicare. You walk in with 2 strikes against you. (Female, Black, participant 6, prior discrimination)

Objective 2: Understand Whether and How Treatment Preferences Are Elicited by Providers and Routinely Incorporated Into Clinical Care for Depression

Theme 2a: Treatment preferences are not often systematically elicited

Most participants could not recall being asked outright about their treatment preferences in health care or discussing them explicitly with providers. More often, study participants reported being offered a menu of treatment options based on what was available at a particular clinic or by a particular provider.

Interviewer:

So does she know, does she know you want talk therapy, or—?

Participant:

We never got that far deep as far, really. Having a talk like this, we never did. Which I think is very important. At least 1 time, you know. (Female, Hispanic/Latino, participant 3, prior discrimination)

Interviewer:

Can you tell me a little bit about how do you communicate that preference to your provider? Is it something you talk to him or her about?

Participant:

No, exactly that. I leave providers that I don't think are a good fit. (Female, White, participant 5, no prior discrimination)

Theme 2b: Regardless of past health care discrimination, patients generally value a trusting relationship with clinicians that facilitates an individualized, fully informed approach to selecting optimal treatments

Whether they were discriminated against or not, participants expected a “good” mental health provider to be someone who listens to and understands them:

It's just their presence and the way they talk to you, the way they present themselves to you. They don't come to me or look like they looking down at me, like they better than me. (Female, Black, participant 6, prior discrimination)

I think a good medical provider is someone who is a good listener, someone who is patient and who is willing to allow the person to make choices instead of taking away their power. (Female, Black, participant 20, no prior discrimination)

Two participants also said that for establishing trust with a provider, it was important that a provider have a good “vibe” or the right “forma de ser” (Spanish for “way of being”), which underscored the extent to which participants perceived good providers not only to display but to truly embody these ideal characteristics:

That's why I continue seeing her. It's a great experience, you know, talking with her because she is the person that I can feel that vibe from that I was talking about before—being open and I feel like I can be, you know, open with her without, you know, feeling edgy about it. (Female, Black, participant 1, no prior discrimination)

These aspects (ie, listening, understanding, connection) were seen as hallmarks of a trusting patient-provider relationship, which was essential to ensuring an individualized treatment approach. Several participants preferred to avoid providers who would “push” medication on them either because they had not spent the time to learn whether this would be optimal for the patient or because of participants' belief that this approach was more lucrative for providers:

I think it's someone who would agree with my concept of getting to know the individual before pushing a medication than you know, having the open discussion over the medication before just pushing it on the individual. (Male, White, participant 8, no prior discrimination)

So, a real doctor that's not out for money that really wants to help somebody and don't choke everybody up with pills. Because that's how they're making their money. (Female, Black, participant 6, prior discrimination)

Participants who had not experienced discrimination were somewhat more likely to describe knowledge and clinical skills as important provider attributes:

Someone who consistently has good outcomes … I like that she makes decisions based on evidence-based medicine. (Female, White, participant 5, no prior discrimination)

An ear to listen, definitely. Knowledge in their practicing area is another, sometimes a good recommendation from someone. (Female, non-Hispanic/Latino Black, participant 11, no prior discrimination)

In contrast, participants who had experienced health care discrimination more often used descriptions such as “sincere,” “reliable,” “honest,” and other phrases that described the provider's intentions:

Well I guess they have to listen and I guess, when it comes to providing advice or feedback or, you want them to, I guess, you want them to be knowledgeable. And you want them to be sincere and you want them to be coming from a place that has your best interests. (Female, White, participant 13, prior discrimination: mental health diagnosis)

So, somebody who is reliable and honest. Like, I've had too many bad experiences. I had this woman and then, like, after a year I was like, this doesn't seem to be going anywhere and she was like, yeah I never understood you, and I'm like, excuse me!? … For a year?! (Female, Black, participant 14, prior discrimination)

One participant with prior discrimination also directly contrasted the value of knowledge and skill vs having respect and care in the patient-provider relationship:

Participant:

He is not the best doctor. He is really kind of crappy. He has no knowledge on how to be a therapist, but he is a good person, and he has known me a long time, and he cares about me. (Female, White, participant 19, prior discrimination: mental health diagnosis)

Objective 3: Understand to What Degree Treatment Preferences Are Shaped by Experiences of Health Care Discrimination

Theme 3a: Regardless of past health care discrimination, treatment preferences are fluid and shaped by SDM with providers whom the patient trusts

Establishing a balance of power within the patient-provider relationship was seen as essential to facilitating an open discussion of preferences and engaging in full SDM for mental health treatment. Providers were seen as being responsive to patient-specific needs when they moved beyond simply describing the array of treatment options and toward fully using SDM if the patient wanted to have a voice in treatment preference:

I have had providers in the past that I didn't feel were customizing my therapy. And sometimes I felt like they were annoyed by questions that I asked about medicines they wanted to put me on and concerns that I had…. I don't know if it was a male-female thing, or just the patriarchy of medicine where doctors … just expected to not be questioned and just, you know, to do what the doctor says no matter what. (Female, White, participant 5, no prior discrimination)

Participants' willingness to try various treatment options (including those outside their originally stated treatment preferences) was heavily contingent on this trusting relationship. One participant considered herself an “antimedicine person” but said she would be willing to discuss medication for her depressed mood after exploring other options:

I think she [provider] noticed I wasn't myself, you know, when they see you over the years or even with the pregnancy when they see you the umpteenth time within the last few months, like almost every week. You know, they kind of get to know your personality…. [Later in interview] I would want to explore other avenues to see what we can possibly do before we go down that avenue or at least give other options. I don't want medicine to ever be just the only option. (Female, Black, participant 11, no prior discrimination)

Given that reaching a mental health diagnosis can take time and that optimal treatment can change over time, narratives revealed that the decision to remain in mental health care also hinged on maintaining this trusting relationship through which SDM could continue to occur. A case in point is the narrative shared by another participant whose trusted primary care provider recognized her depression and successfully leveraged their established relationship to encourage her to try medications, despite her initial resistance. The medications gave her enough motivation to pursue the arduous task of finding a counselor despite being uninsured at the time. When asked what might have happened had she not sought and ultimately received counseling, her response was blunt: “At the time, I was suicidal. So, I would probably be dead.” (Female, White, participant 10, prior discrimination: income/social class)

Theme 3b: Patients who have experienced health care discrimination face greater challenges to forming trusting relationships with providers to engage in SDM, which both elicits and shapes preferences

A strong patient-provider relationship was important for all respondents, but for those who had experienced health care discrimination, a trusting relationship with a provider was often critical for more than just establishing an appropriate care plan. For those experiencing discrimination or stigma based on their mental health diagnosis, medical providers were often able to restore agency to someone who was otherwise vulnerable to losing it:

One of the most important things [about my doctor] is that he'll write letters … to people that I need help from or that I'm getting screwed over by … I get, I buy a little bit of, it's like, leverage in society. Because I feel like by myself with the label I have, people are just going to walk all over me, but I can have him write a letter and it lifts me up, protects me. (Female, White, participant 19, prior discrimination: mental health diagnosis)

Issues of trust and power imbalance were often exacerbated when participants faced multiple forms of discrimination, including outside health care settings. When these experiences with discrimination were not explicitly addressed, true SDM for treatment preferences could not occur.

I was discriminated against, kind of like, I was told that no, you have quite the mental illness and you need, if you ever want to go to school or get a boyfriend or something, you need medicine. You are going to be on the medicine ‘til your 40s or 50s, what you have is chronic … and that's just the psychiatrist's opinion thinking that if someone has really bad mental illness that they need medicine. (Female, White, participant 15, prior discrimination: mental health diagnosis)

Interviewer:

Okay. And while in this case if you talk directly with your doctor or the nurse that sees you about these preferences, would you be comfortable to talk to face-to-face?

Participant:

Yeah, it depends on who they are and their energy. (Female, Black, participant 6, prior discrimination)

Although most participants reported difficulty finding a medical provider who was a “good fit,” not all participants felt equal agency to change their provider and/or treatments that were not working for them. Below we highlight sample quotations demonstrating high vs low agency to advocate for different treatment and/or providers, contrasting differences in agency between those who did and did not experience health care discrimination:

1. High agency in response to poor provider or treatment fit: no prior health care discrimination

   Participant:

   I leave providers that I don't think are a good fit.

   Interviewer:

   And how often do you encounter providers that you don't think are a good match?

   Interviewer:

   About half the time. (Female, White, participant 5, no prior discrimination)

   Interviewer:

   So if your provider all of a sudden stopped with talk therapy and says, okay, we only have medication, would you move to another [provider]?

   Participant:

   I would. (Female, Black, participant 20, no prior discrimination)

   If I feel like you're not giving us the best care. If you're not listening to our issues or if we have an issue and I don't feel like it's the best care I get a second opinion, I'm out. I just—I can't. (Female, Black, participant 11, no prior discrimination)

   I say, well, if you're not happy with your provider, you can always advocate for yourself and ask for a different provider. (Female, White, participant 16, no prior discrimination)

   Interviewer:

   What should happen if your provider or clinic stopped offering the treatments that you like?

   Participant:

   I would find somewhere else to go. [Laughter]

   Interviewer:

   You could change providers?

   Participant:

   Yeah, yeah, I would change my provider, yeah. (Female, Black, participant 17, no prior discrimination)

2. Low agency in response to poor provider or treatment fit: prior health care discrimination
   You know but it's—really they got—we don't have a choice who our doctors are sometimes. (Female, Black, participant 6, prior discrimination)

   Interviewer:

   So if your current provider stopped, like, having medication, then would you switch to another provider? And what about talk therapy?

   Participant:

   Probably.

   Participant [later in interview]:

   You kind of have to be a bit submissive. (Female, White, participant 13, prior discrimination: mental health diagnosis)

The need to establish a trusting patient-provider relationship through which patients could obtain their treatment of choice was described in more charged language by patients who had previously experienced health care discrimination. In some cases, narratives demonstrated that these participants needed to work harder to find the right provider or the right treatment: “It doesn't matter to me to go a little farther if it's someone who is really worth going farther for.” (Female, Hispanic/Latino, participant 9, prior discrimination)

Because all of my emotional upset is being caused by this distorted thought … so I have found CBT [cognitive behavior therapy] to be and EMDR [eye movement desensitization and reprocessing] to be 2 of the most helpful things and so I have to have someone who, even if it is not their orientation, they're familiar with it and they are going to be actively engaged. Because I can slide into, you know, a very, very, very dark and hopeless place and I haven't for years because of the work that I have done and then the people I rely on as providers being actively engaged in listening, deep listeners, deep listening. (Female, Black, participant 14, prior discrimination)

For patients who had experienced discrimination, and especially for those experiencing discrimination because of a mental health diagnosis, a trusting relationship with a provider was the pathway through which they could exercise agency. If those trusted providers were to remove their patients' agency (eg, by insisting on prescribing medications for the patient against preferences they had established over time with their providers), several participants described the feelings of fear, disappointment, and/or betrayal that would ensue:

Well, he already asks me, you know, this medication will be good or that one will be good, but that's okay—he can say it. But if he starts saying, “No, you need this,” I'll be like running for the hills. (Female, White, participant 19, prior discrimination: mental health diagnosis)

We will be alienated from each other, I'd be frightened and scared and feel alone…. I mean, it would ruin our whole relation, everything we've built for 15 years. (Female, White, participant 19, prior discrimination: mental health diagnosis)⁷¹

These results highlight the challenges to receiving mental health treatment, which may be exacerbated by experiences of discrimination. Treatment preferences are not always systematically elicited, and ultimately preferences are malleable, especially in the presence of a strong, trusting relationship with clinicians. Past experiences of health care discrimination pose a significant challenge to establishing trusting relationships with a provider and engaging in SDM.

Type 2 Diabetes Sample Thematic Results

Key themes elicited from interviews with patients with type 2 diabetes yielded similar results to the depression sample. Overall themes for the diabetes sample focus primarily on discrimination experiences by race/color.

One participant, who identified as Brown, described an instance where he felt he had been discriminated against because another patient entered the waiting room who had a lighter skin tone than him, thereby introducing within the interview itself the point that discrimination experiences often range by skin tone rather than race or ethnicity alone:

Participant:

So when I stepped off, this other person comes in, different shade of color than me, and the person just lifted her head, how can I help you and I got mad, and I definitely said something, yeah, I said some words.

Interviewer:

Do you mind me asking was that other person White or just they weren't Black, the other patient who came in?

Participant:

I'm not Black, I'm Brown.

Interviewer:

Okay. They weren't Brown?

Participant:

No, they weren't.

Interviewer:

Were they White?

Participant:

They were closer to White. (Male, Brown, participant 35, prior discrimination)

The racially motivated discrimination experiences that participants reported from their interactions with their provider or the provider's front-desk staff spanned multiple domains but generally included negative assumptions about health behaviors, such as risky sexual activity, lack of motivation to diet, and propensity for substance abuse.

For example, 1 participant reported that a medical provider insisted on unjustified HIV testing and made condescending assumptions about her ability to diet because of her race:

This person said to me that I was at high risk because of having unprotected sex with my boyfriend that I had been with for 7 years, because I didn't know where he had been. (Female, Black, participant 38)

In another example from the same participant, a provider reacted to what the participant described as her simple joke that she could not resist chocolate with stern, paternalistic language: “If you want to stay fat … ” (Female, Black, participant 38)

Another participant reported biased perceptions of her propensity to abuse medications and/or illicit substances based on her race and the fact that this bias made her less likely to bring up her own pain and suffering with her doctor:

You know, it's just like sometimes you get discriminated because of your color…. I don't want to play the race game, but it just seems to me like … you get treated a little bit different … if I'm in extreme pain, if I ask for [pain medications] I feel like I won't get it…. So it make you not want to ask even though you're in pain and you suffer. (Female, Black, participant 27)

In addition to these negative assumptions about health behaviors, 1 participant highlighted their provider's misguided assumptions of newborn jaundice risk based on assumptions related to the race/color of both parents of the child. As a result, the health care provider assumed that the skin tone of the participants' newborn was indicative of severe jaundice risk without considering a different possibility—that both parents may not be White—and so made the mother believe the newborn's life was in danger, almost prompting more serious medical intervention even though the baby was actually fine. The participant recounted this as a highly traumatic experience (Female, White, participant 39).

Participants recounted a range of reactions to experiences of discrimination. Some participants had internalized negative emotions, while others disengaged from medical care (they reported leaving the clinic or office and not returning to that clinic for care or avoiding care entirely). As in the depression sample, only a few individuals reported formally reporting the incident at the clinic. For example, 2 participants who shared stories in interviews about experiencing inappropriate HIV testing described feeling offended:

I was upset…. I took offense. (Female, Black, participant 38)

That was really an insult. (Female, Black, participant 36)

Beyond feeling upset, another participant received inappropriately judgmental comments that made assumptions about her sexual activity. While administering the pregnancy test, the doctor asked about her marriage status and implied that the participant should not be having a child outside of a marriage “unless you've been raped.” She shared her decision to report the incident, leave the office, and never return:

So when I left that office, I was so upset I said I will not ever, ever, ever go to any doctor ever again that treats a patient that way … of course this made me to file the complaint with the Medical Board. (Female, Black, participant 29)

These reactions begin to illustrate potential pathways of how experiences of discrimination affect patient care among individuals with diabetes, such as reducing the quality of the patient-provider therapeutic alliance; neglecting, misdiagnosing, or overlooking a serious medical issue; patients feeling forced to leave the clinic or delaying/avoiding needed care; and patients experiencing decreased satisfaction. Importantly, people with diabetes reported experiencing health care discrimination throughout a multitude of their health care interactions, not only in the course of diabetes care.

When asked to describe whether or how taking a survey such as the one administered in aim 1 (which included measurements of preferences as well as reports of prior discrimination) would be helpful, participants hypothesized several mechanisms by which such data collection as part of patient care might be helpful in providing patients with the best care for their diabetes treatment:

You have to sit down and ponder, like, is this the right choice, you know, is the one—something like that make you think that you're getting a better treatment, the best you could do for yourself. (Female, Black, participant 27)

Interviewer:

Do you think that answering some of these other questions as well around like what you think makes a good doctor or whether you were treated unfairly in the past? Do you think that would be helpful if your doctor knew that about you going into your appointment?

Participant:

Yes I think it would be, yes.

Interviewer:

Okay. Are there any parts of it that you think would be most helpful or kind of that were most important to you or interesting when you're filling [the survey] out?

Participant:

Well, I think it would be most important that they are aware of my concerns and my perspective. (Male, Black, participant 30)

Aim 3

Interview clinician and health care administration stakeholders to (1) assess how providers typically elicit and incorporate information about patient treatment preferences and prior discrimination into treatment plans and (2) determine the potential utility of collecting more structured data to improve the tailoring of treatment plans.

Theoretical Framework

Aim 3 consisted of a qualitative study using semistructured interviews of clinical care providers in an integrated safety-net health care system that serves a racially and ethnically diverse patient population. This study was initiated as part of a broader effort to establish more effective methods for developing treatment plans for depression and diabetes and also potentially reduce known racial and ethnic disparities in treatment for these conditions.

In aim 3, we used what we had learned in aims 1 and 2 to ask providers during semistructured qualitative interviews about how they learn patients' preferences for treatment of depression and diabetes; how they learn about patients' prior experiences of discrimination; and whether data from a survey, such as that described in aim 1, could be implemented into their clinical practice to better tailor treatment plans with patients. The goal of conducting this work was to facilitate providing specific guidelines to incorporate a patient's past experiences of health care discrimination into the elicitation of patient preferences and thereby to improve the tailoring of treatment plans.

We purposely selected this setting for the study because most outpatient treatment of depression and diabetes occurs in primary care settings and because a major focus of our study is on optimizing treatment among racially/ethnically diverse groups. Because little is known about these areas and because these factors are nuanced and complex, we chose an exploratory qualitative approach (see McCormick et al in the Related Publications section).

Qualitative Data Collection

Four trained members of the study team conducted and coded the 20 interviews, which were split in half and coded independently by 2 of the 4 coders (ie, 10 interviews per 2 coders). The coding process was performed using Dedoose qualitative analysis software.⁶⁰ Coders also generated a summary analytic memo after coding each interview. The coding process consisted of a thematic analysis and was performed using a qualitative code “tree” generated deductively by the senior qualitative researcher per the domains included in the interview. The qualitative code tree (Appendix D) included codes with up to 3 levels, which were generated as open codes (inductively) early in the analysis during qualitative training sessions led by the team's senior qualitative researcher. The initial coding sessions allowed for further refinement of the code tree early on by identifying and adding new themes and relevant codes that emerged from the interviews but were not included in the interview guide. Results were ultimately organized into 5 overarching themes. Finally, upon completion of the coding process, pairs of researchers met to review and compare the codes they had applied to identify and resolve discrepancies. When discrepancies had been resolved, codes generated by the 2 researchers were consolidated.

Sample

Table 5 describes the study participant characteristics. A total of 20 semistructured qualitative interviews were conducted with 12 primary care providers (physicians and primary care nurse practitioners), 3 primary care leaders or management specialists (clinic directors, managers, integration specialists), 3 social workers or case managers, and 2 pharmacists. Ten (50%) participants identified as female, 17 (85%) identified as White, and 10 (50%) reported proficiency in a language other than English. To recruit participants, we sent an email to the practice manager and staff with clinical roles in which we briefly described the primary aims of the study and offered a $75 gift card for participation. Providers interested in participating were asked to email the project's research coordinator to schedule an interview. All contacted participants agreed to participate and were recruited into the study.

Table 5

Study Participant Characteristics.

Four trained interviewers conducted the 20 interviews in person between December 6, 2018, and January 23, 2019. All participants provided written informed consent to participate in the interviews, which were audio-recorded and professionally transcribed. The interview lasted approximately 1 hour and covered 7 content domains (Appendix E): (1) interviewee's role and responsibilities at the primary care clinic; (2) how the interviewee obtains information from patients about barriers to treatment and treatment preferences; (3) interviewee's conversations with patients about experiences of discrimination in health care settings; (4) interviewee's views about how past health care discrimination influences treatment preferences; (5) interviewee's ideas about how elicitation of information on barriers, treatment preferences, and past health care discrimination could improve treatment plans; (6) interviewee's ideas about how to collect information on barriers, treatment preferences, and past health care; and (7) perceived feasibility of using a DCE approach to assess treatment preferences. This study focused only on results from domains 2, 3, and 5.

A key objective of this study was to understand providers' views of the relationship between past experiences of health care discrimination and preferences for treatment as well as whether and how discrimination may affect expressed preferences. Thus, in the study interview, participants were presented with patient data obtained from these earlier phases of the study and asked for their reactions. In particular, data showing the high prevalence of perceived discrimination among racially/ethnically diverse people with depression was presented, as well as qualitative data indicating that experiences of discrimination among this population often lead to disengagement from medical care (see interview guide in Appendix E). Because we acknowledged that providers may not be the best historians regarding their own role in discrimination, the stakeholder team determined that asking providers directly about their own role in discrimination toward patients would be less solution oriented than asking implementation-focused questions about how data on discrimination experiences would be useful for providers; we also sought feedback on what kinds of supports or training providers would need to use this data in practice. Participants were presented with data showing that patients with diabetes who experienced discrimination were more likely to prefer behavior modification (diet/exercise) over medications, and patients with depression who had experienced discrimination were more likely to prefer medications over talk therapy.⁴¹

Results

Theme 1: Providers obtain information about patient preferences during treatment planning

Most providers described using an SDM process in which they offer different treatment options, and then evaluate the patients' response to those options before finally reaching a consensus on an optimal treatment choice. Another common approach articulated by providers was to use knowledge of the patients' past preferences and values gained through long-term primary care relationships to inform present decision-making about treatment options or to use trust built up over time to encourage honest sharing of treatment preferences:

I think, when I'm developing a plan with a patient, I develop it with the patient so at the point of [decision-making] … I generally talk about different treatment options, and ask the patient what their preferences are and how they want to proceed.

However, providers also frequently mentioned employing indirect methods of ascertaining patient preferences, such as waiting for the patient to volunteer their preferences (vs asking the patient directly) or observing the patient for nonverbal cues to indicate treatment preferences: “Well, I think that you watch body language, you kind of listen to people's cues, verbal, nonverbal, and kind of respond to that.”

Many providers also described the need to overcome barriers to eliciting treatment preferences in clinical encounters. Common barriers mentioned by providers included patients' low educational or health literacy level and limited English-language proficiency, which can adversely affect communication, even when using a translator. Providers often mentioned that such barriers prevent patients' full understanding of treatment options and providers' understanding of patients' preferences:

Sometimes language. It's hard, language and culture are definitely things I try to be conscious of, like literacy level and medical education–level kind of stuff, it's clearer that people who have more resources and more language ability or medical literacy will be able to explain what they're interested in more and so sometimes in the situation which those present some level of kind of barrier to care, then it can be a little harder to understand what they want.

Theme 2: Providers rarely elicit narratives about past experiences of discrimination

Most providers indicated that they do not initiate conversations with their patients about past experiences of discrimination in health care settings. When such conversations occur, they are usually triggered by patients describing past negative experiences with other providers, usually outside the study institution:

I don't blankly ask about it very much, but if I get some smell of something, I'll actually push pretty hard to understand if something like that's happened. But I would love to kind of do a little more prompting and screening of whether someone feels discrimination.

In addition, most providers indicate that patients also do not typically initiate conversations about discrimination but rather that they may come up indirectly through conversations about other topics, such as resistance to a particular treatment or when patients are asked about past general experiences with other providers (not specific to discrimination):

It doesn't really come up, that people feel like they've been discriminated [against] in care. It doesn't. I mean, we know that it's out there, for various groups of folks. But, it doesn't come up. I've never had someone describe to me, an incident where they felt like they were discriminated against.

Theme 3: Provider reactions were mixed regarding their willingness to talk about discrimination in the health care setting

Most providers believed that eliciting information about their patients' past experiences of discrimination in addition to information about treatment preferences would be useful in optimizing treatment plans and, furthermore, that they would be willing to do this. Some providers, however, had mixed feelings about eliciting information, and a few felt that it would definitely not be helpful.

Providers' commonly cited reasons for supporting the idea of eliciting past discrimination experiences included that it would enable them to be more sensitive and responsive to their patients' needs and that it could open the door to other difficult but useful conversations to provide insight into their (or their institution's) role in contributing to experiences of discrimination and enhance patient-provider trust:

I mean, I think it's really important to know that history. And I feel like, at the very least, acknowledging, and knowing that that's happened would go a long way to create a certain level of trust and common understanding. I think the real question is the how. As with everything, in theory … I think it could be incredibly helpful.

Providers who commonly cited reasons for skepticism about elicitation of prior discrimination experiences stated that this could be perceived by patients as invasive and thus provoke feelings of discomfort and could add additional time burdens to already overburdened clinic schedules, particularly in the context of unproven benefits of collecting such information. Several providers were also concerned that collecting such information could be uncomfortable for the provider because they may not know how to address patients' feelings of past discrimination, noting that providers are not typically trained for this:

In practice, how to do it in the context of a busy clinic, with competing demands, really … And it feels like, that's actually the start of a larger conversation. A conversation I don't have time to have.

Some providers did not see value in eliciting narratives of past discrimination, even apart from the time constraints:

It wouldn't change how. I try to do the same thing with all my patients. If someone tells me that they were treated poorly because they didn't understand in a system, vs another patient, I'd treat them the same way. Again, that's another reason why I don't see those conversations being super-productive for me, because I'm treating them the same way. I would hope to be treating this patient the same way, so regardless if they told me that they had that bad experience before.

Some but not all providers said that there was utility in collecting information on preferences and past experiences of discrimination. Half of the providers felt that the general concept of collecting information on both patients' preferences—what they value most in thinking about treatments (eg, avoidance of adverse effects, low cost, number of clinic visits)—and patients' past experiences of discrimination through a “previsit” survey could be useful in informing discussions about and formulating treatment plans in outpatient practice. Some providers felt that such a process could help make the approach to collecting preferences more systematic, start or guide a conversation about treatment, and help set goals of treatment. Others saw potential value in gathering this kind of information in a previsit survey because it could enhance confidence in the provider or institution (eg, it demonstrates their care and concern about patient preferences and past negative health care experiences).

These factors led some providers to believe that such a process could, in turn, lead to more complete SDM and thus to treatment plans that are best matched to patient preferences and most likely to be adhered to in the long term:

I think it is helpful. I feel like I get at that through a conversation, but imagine having this up front, prior to that conversation, to be able to say, “I see that you really prefer talk therapy over medication therapy. And in your context, when you're severely depressed, I really hear you. But I also would recommend …,” I think that would come through better to the patient, knowing that I already had listened to that preference.

Several providers, however, expressed doubt that such an approach would be useful, commenting that it is simply more useful to directly talk with patients about these issues (rather than use a survey) and that they already adequately assess and incorporate patient preferences into treatment decisions. Other causes of skepticism included a concern that providers might allow the survey to substitute, in part, for direct conversation, which could lead to both less accurate information and diminished trust in the patient-provider relationship.

Theme 4: Significant practical barriers exist to systematic data collection pertaining to preferences and health care discrimination

Despite general support for the idea of collecting information about preferences and discrimination through a previsit survey, most providers articulated a range of concerns about whether and how a previsit survey might be practically implemented within their clinic and health care system. One commonly cited concern was the potential to further burden already overburdened clinic workers, causing staff and provider fatigue and distracting them from important competing demands. Many providers also cited concerns about the extra burden on patients, who also have competing time demands and medical concerns when visiting the clinic. Another anticipated challenge was whether it was possible to write survey questions that would be understandable to a wide range of patients with different levels of health and general literacy, educational attainment, cognitive capacities, and languages spoken. In addition, many providers anticipated significant logistical challenges in implementing such a survey, with specific concerns voiced about where (home vs clinic) and how (paper vs electronic format) the survey might be delivered to patients and how this choice would affect clinic workflows. Finally, many providers articulated concerns about the confidentiality of the type of sensitive information that would potentially be collected in such a survey. Some providers also felt that patients' concerns about confidentiality might inhibit them from being truthful in answering survey questions, thus diminishing the survey's value.

Provider Guidelines

Upon completion of data collection and analysis of aims 1 through 3, we synthesized findings into a preliminary set of provider guidelines. These guidelines were crafted with the assistance of members of the research team who have clinical expertise. They have been disseminated to other clinicians at CHA and are used as a discussion starter for educational activities conducted by our partner center, the Center for Health Equity and Education at CHA. Findings from all 3 aims inform these recommendations, with a focus on amending and expanding upon a well-known template of SDM. The full text of these guidelines is available in Appendix F.

Discussion

Our study challenges the current prevailing wisdom about racial/ethnic differences in preferences for depression treatment. Prior studies have posited that Black and Latino patients prefer talk therapy over medication to treat depression; results from these studies are frequently cited and are often used to explain the prevalence of and solutions to wide racial/ethnic disparities in access to mental health care. In contrast, our study posits that while White patients, on average, prefer medication to talk therapy, Black and Hispanic/Latino patients do not have a significant preference for 1 type of treatment over another. This is an important finding with implications for solutions to reducing access disparities. Possible explanations for this discrepancy are that our study was a community-based sample as opposed to the primary care–based sample used in prior studies and because the preferences adjusted for other barriers to treatment, such as travel time, cost, and provider language—all barriers to talk therapy—are disproportionately endorsed by racially/ethnically diverse groups.

We build upon this important finding by identifying that past experiences of discrimination influence depression treatment preferences and that Black and Hispanic/Latino patients who have experienced discrimination shift to preferring medication over talk therapy, even after adjustment for the barriers mentioned previously.⁴¹

Likewise, in the diabetes sample, White and Black respondents who reported vicarious discrimination shifted preferences toward behavior modification; this choice did not shift preferences among Hispanic/Latino respondents. Both of these results were true for men but only marginally true for women (see Creedon et al in the Related Publications section).

These findings endorse the conceptual model that we sought to test in this research project—namely, that patient preferences are heavily influenced by prior negative encounters, both those experienced personally and by family and friends, in the health care setting. This finding has ramifications for SDM during treatment planning and during the maintenance phase of treatment. Eliciting preferences for treatment without understanding the prior experiences of patients and their families or friends may miss a component that could be central to improving patient-centered care.

This outcome is particularly relevant for efforts to reduce health care disparities, as defined by the National Academy of Medicine (formerly the Institute of Medicine).¹³ Under the National Academy of Medicine definition, treatment differences are “allowable” if they result from patient preferences. However, if—as suggested by the present findings—preferences are in part derived from past discrimination (an “unallowable” source of treatment differences under the National Academy of Medicine definition) as well as from compromised SDM, then untangling the relationship between past experiences of health care discrimination and current preferences will be critical to addressing disparities.

Our results show that experiences of health care discrimination, either personal or vicarious, are highly prevalent among people in racially/ethnically diverse groups living with depression and diabetes and that these experiences have an impact on treatment preferences for depression and diabetes. These associations are further modified by language and gender. We acknowledge significant variation in racial presentation among Hispanic/Latino individuals who may not experience discrimination because of perceived racial identity but may be discriminated against for other factors, such as accented speech. This finding is congruent with previous research, which identified unique themes regarding gender values and health behaviors among a sample of Black and Hispanic/Latino men with diabetes. Participants expressed

[that] the need to maintain a strong image to the outside world, and the need to maintain control of themselves served as barriers to seeking out and engaging in diabetes self-management behaviors,

which likely differ significantly compared with how gender values are embodied by women.⁶¹

Our quantitative LCA study highlights the intersectional nature of health care discrimination and its disproportionate impact on mental health (see Adams et al in the Related Publications section). To our knowledge, this study is the first to disentangle the impact of identity-level attributes on perceptions of health care discrimination using a latent-variable approach. Our findings affirm the diverse nature of discrimination experiences across intersectional attributes, both personally and vicariously, and its joint effect on mental health. By creating population-level discrimination subgroups across multiple axes of identity (eg, race; ethnicity; language; lesbian, gay, bisexual, transgender, queer status) and perspectives (ie, personal and vicarious), our study finds that higher instances of racial discrimination as well as moderate levels of intersectional discrimination have adverse effects on mental well-being. Approaches to mitigate the deleterious effects of health care discrimination should consider innovative approaches to adequately address the social structures that perpetuate health and health care inequities. To provide more equitable care, health care organizations are also uniquely positioned to consider multilevel quality improvement programs that address individual patient experiences; interpersonal exchanges among providers, staff, and patients; and reshaping organizational culture to reduce instances of discrimination in the clinical environment.

In the quantitative results examining the mediation of therapeutic relationships in the association between health care discrimination and type 2 diabetes care quality, we found that the therapeutic relationship was a significant mediator of the relationship between health care discrimination and diabetes care quality (see Flores et al in the Related Publications section). The results suggest that the therapeutic relationship is a critical intermediary mechanism in our hypothesized model. We found that the negative association between perceived health care discrimination and ratings of type 2 diabetes care quality may not only operate in a direct pathway but also may be shaped by discriminatory treatment within the patient-provider relationship. In other words, when patients experience unfair treatment in the health care system, particularly because of differences in race/ethnicity or language, it diminishes care quality ratings by compromising the belief that the provider has a strong therapeutic interest in the patient's care.

Qualitative research with participants who have depressive symptoms or type 2 diabetes reinforces the impact that prior experiences of health care discrimination have on engagement in treatment and how patients respond when asked about their preferences for treatment. While the interviews with participants with depression highlighted the important role that discrimination and stigma based on mental illness can play, the interviews with participants with diabetes demonstrated that participants in our sample actually are on the receiving end of a broad range of experiences of discrimination throughout their clinical histories. Participants with diabetes faced judgments from providers about their ability or interest in following through with a diet regimen. Moreover, participants reported that this theme of dismissing participants' ability to follow specific health recommendations or healthy behaviors carried into other settings. For example, several Black women described instances in which providers ordered inappropriate HIV testing and made assumptions about their or their partners' participation in risky sexual activity. Importantly, these qualitative interviews did more than just reinforce prior quantitative findings; they uncovered how these negative associations can be overcome in the presence of trust, communication, and respect between provider and patient. For people in racially/ethnically diverse groups and those experiencing discrimination based on other aspects of their appearance—such as weight and attire—and/or their mental illness, factors such as trust, communication, and respect between the provider and patient are key to developing rapport, overcoming past experiences of health care discrimination, and improving SDM in treatment planning. Participants in both samples rarely reported their discrimination experiences to managers or institutional leadership. At times, participants reported reactions such as completely disengaging from care with that provider or providers in general. This course of action demonstrates how single or salient experiences of discrimination, as well as the aggregate experience of repeated offenses over time, can have immediate and reverberating impacts on patient care. In the short term, an individual may leave an appointment; in the long term, that individual may avoid seeking care altogether.

Taken as a whole, our qualitative findings suggest the need to integrate 2 other seemingly important factors that shape patient treatment preferences: (1) trusting patient-provider relationships and (2) SDM. Within this revision of our initial model, these 2 factors are recursively related to one another (Figure 6).

Figure 6

Revised Conceptual Model of Adaptive Preferences Informed by Social-Ecological Model.

Conclusions

In a nationally representative quantitative survey, we found that non-Hispanic/Latino Black and Hispanic/Latino respondents did not have a significant preference between talk therapy and medications for depression treatment or between medication and behavioral modification for diabetes treatment. However, these preferences were different if an individual had experienced prior health care discrimination.

Different treatment preferences for these conditions were explored in qualitative interviews in which participants shared narratives that clarify possible pathways to health disparities. Even when participants reported an a priori preferred treatment method, we found that they relied heavily on a trusting relationship with their provider through which to explore possible treatment options and select the best treatment. We found, therefore, that preference elicitation was important not only as a form of measuring patient preferences accurately, especially for racially/ethnically diverse groups, but indeed as a process by which to collaboratively generate such preferences. Experiencing health care discrimination made it more difficult for participants to find providers with whom they felt they could engage in trustworthy, open SDM and therefore could handicap both the process of preference elicitation and the formation of preferences.

In interviews with clinicians, we found that while clinicians were concerned about the impact of health care discrimination on patient trust in providers, they did not routinely have conversations about discrimination in the course of patient care, both because patients did not routinely volunteer this information and because providers did not routinely ask for it. Because providers reported many different processes for eliciting patient preferences, some of which involved more deliberate questions about patients' prior experiences with treatment and others that did not, we conclude that SDM for preferences, especially with patients from marginalized backgrounds and/or who have been stigmatized because of a mental or physical health disorder, can be improved by more intentionally incorporating a consideration for prior negative experiences, including discrimination experiences.

References

1.

American Psychiatric Association. Practice Guideline for the Treatment of Patients With Major Depressive Disorder. 3rd ed. American Psychiatric Association; 2010. Accessed November 6, 2020. https:​//psychiatryonline​.org/pb/assets/raw​/sitewide/practice_guidelines​/guidelines/mdd.pdf

2.

Gelhorn H, Sexton C, Classi P. Patient preferences for treatment of major depressive disorder and the impact on health outcomes: a systematic review. Prim Care Companion CNS Disord. 2011;13(5):PCC.11r01161. doi:10.4088/PCC.11r01161 [PMC free article: PMC3267514] [PubMed: 22295273] [CrossRef]

3.

Dwight-Johnson M, Unutzer J, Sherbourne C, Tang L, Wells KB. Can quality improvement programs for depression in primary care address patient preferences for treatment? Med Care. 2001;39(9):934-944. [PubMed: 11502951]

4.

Swift JK, Callahan JL. The impact of client treatment preferences on outcome: a meta-analysis. J Clin Psychol. 2009;65(4):368-381. [PubMed: 19226606]

5.

Armstrong K, Hughes-Halbert C, Asch DA. Patient preferences can be misleading as explanations for racial disparities in health care. Arch Intern Med. 2006;166(9):950-954. [PubMed: 16682567]

6.

Institute of Medicine. Unequal treatment: what healthcare providers need to know about racial and ethnic disparities in health-care. Published March 2002. Accessed February 1, 2021. https://www​.nap.edu/resource​/12875/disparities_providers.pdf

7.

Alegria M, Chatterji P, Wells K, et al. Disparity in depression treatment among racial and ethnic minority populations in the United States. Psychiatr Serv. 2008;59(11):1264-1272. [PMC free article: PMC2668139] [PubMed: 18971402]

8.

National Public Radio, Robert Wood Johnson Foundation, and Harvard T.H. Chan School of Public Health. Discrimination in America: Experiences and Views of Latinos. Robert Wood Johnson Foundation; 2017. Accessed November 6, 2020. https://www​.rwjf.org​/content/dam/farm/reports​/surveys_and_polls/2017/rwjf441402

9.

National Public Radio, Robert Wood Johnson Foundation, and Harvard T.H. Chan School of Public Health. Discrimination in America: Experiences and Views of African Americans. Robert Wood Johnson Foundation; 2017. Accessed November 6, 2020. https://www​.rwjf.org​/content/dam/farm/reports​/reports/2017/rwjf441128

10.

Alegria M, Chatterji P, Wells K, et al. Disparity in depression treatment among racial and ethnic minority populations in the United States. Psychiatr Serv. 2008;59(11):1264-1272. [PMC free article: PMC2668139] [PubMed: 18971402]

11.

Institute of Medicine. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. National Academies Press; 2003. Accessed February 1, 2021. https://www​.nap.edu/catalog​/12875/unequal-treatment-confronting-racial-and-ethnic-disparities-in-health-care [PubMed: 25032386]

12.

Lin H-C, Erickson SR, Balkrishnan R. Physician prescribing patterns of innovative antidepressants in the United States: the case of MDD patients 1993–2007. Int J Psychiatry Med. 2011;42(4):353-368. [PubMed: 22530398]

13.

Jung K, Lim D, Shi Y. Racial-ethnic disparities in use of antidepressants in private coverage: implications for the Affordable Care Act. Psychiatr Serv. 2014;65(9):1140-1146. [PubMed: 24828481]

14.

Olfson M, Marcus SC. National patterns in antidepressant medication treatment. Arch Gen Psychiatry. 2009;66(8):848-856. [PubMed: 19652124]

15.

Quiñones AR, Thielke SM, Beaver KA, Trivedi RB, Williams EC, Fan VS. Racial and ethnic differences in receipt of antidepressants and psychotherapy by veterans with chronic depression. Psychiatr Serv. 2014;65(2):193-200. [PMC free article: PMC7108428] [PubMed: 24178411]

16.

Williams DR, Mohammed SA. Discrimination and racial disparities in health: evidence and needed research. J Behav Med. 2009;32(1):20-47. [PMC free article: PMC2821669] [PubMed: 19030981]

17.

Balfour PC, Rodriguez CJ, Ferdinand KC. The role of hypertension in race-ethnic disparities in cardiovascular disease. Curr Cardiovasc Risk Rep. 2015;9(4):18. [PMC free article: PMC4577235] [PubMed: 26401192]

18.

Hunt LM, Truesdell ND, Kreiner MJ. Genes, race, and culture in clinical care: racial profiling in the management of chronic illness. Med Anthropol Q. 2013;27(2):253-271. [PMC free article: PMC4362784] [PubMed: 23804331]

19.

Cabana MD, Lara M, Shannon J. Racial and ethnic disparities in the quality of asthma care. Chest. 2007;132(Suppl 5):810S-817S. [PubMed: 17998345]

20.

Alegria M, Chatterji P, Wells K, et al. Disparity in depression treatment among racial and ethnic minority populations in the United States. Psychiatr Serv. 2008;59(11):1264-1272. [PMC free article: PMC2668139] [PubMed: 18971402]

21.

Heisler M, Smith DM, Hayward RA, Krein SL, Kerr EA. Racial disparities in diabetes care processes, outcomes, and treatment intensity. Med Care. 2003;41(11):1221-1232. [PubMed: 14583685]

22.

Dwight-Johnson M, Sherbourne CD, Liao D, Wells KB. Treatment preferences among depressed primary care patients. J Gen Intern Med. 2000;15(8):527-534. [PMC free article: PMC1495573] [PubMed: 10940143]

23.

Katz JN. Patient preferences and health disparities. JAMA. 2001;286(12):1506-1509. [PubMed: 11572745]

24.

Collins TC, Clark JA, Petersen LA, Kressin NR. Racial differences in how patients perceive physician communication regarding cardiac testing. Med Care. 2002;40(1):I27-I34. [PubMed: 11789628]

25.

Cooper LA, Gonzales JJ, Gallo JJ, et al. The acceptability of treatment for depression among African-American, Hispanic, and White primary care patients. Med Care. 2003;41(4):479-489. [PubMed: 12665712]

26.

Houle J, Vilalge B, Bealieu M, Lesperance F, Rondeau G, Lambert J. Treatment preferences in patients with first episode depression. J Affect Disord. 2013;147(1-3):94-100. [PubMed: 23167975]

27.

Gum A, Areán P, Hunkeler E, et al. Depression treatment preferences in older primary care patients. Gerontologist. 2006;46(1):14-22. [PubMed: 16452280]

28.

Huang ES, Brown SES, Thakur N, et al. Racial/ethnic differences in concerns about current and future medications among patients with type 2 diabetes. Diabetes Care. 2009;32(2):311-316. [PMC free article: PMC2628700] [PubMed: 19017766]

29.

Lipton RB, Losey LM, Giachello A, Mendez J, Girotti MH. Attitudes and issues in treating Latino patients with type 2 diabetes: views of healthcare providers. Diabetes Educ. 1998;24(1):67-71. [PubMed: 9526327]

30.

de Weerdt I, Visser AP, Kok G, van der Veen EA. Determinants of active self-care behaviour of insulin treated patients with diabetes: implications for diabetes education. Soc Sci Med. 1990;30(5):605-615. [PubMed: 2408154]

31.

Chatterji P, Joo H, Lahiri K. Racial/ethnic- and education-related disparities in the control of risk factors for cardiovascular disease among individuals with diabetes. Diabetes Care. 2012;35(2):305-312. [PMC free article: PMC3263918] [PubMed: 22190677]

32.

National Center for Chronic Disease Prevention and Health Promotion. National diabetes fact sheet 2011. US Department of Health and Human Services, Centers for Disease Control and Prevention; 2011. https://www.cdc.gov/diabetes/pubs/pdf/methods11.pdf

33.

Muhuri PK, Machlin SR. Statistical Brief #518: Treatment and Monitoring of Adults With Diagnosed Diabetes by Race/Ethnicity, 2015-2016. Agency for Healthcare Research and Quality; 2018. [PMC free article: PMC541162] [PubMed: 31083860]

34.

Piette JD, Bibbins-Domingo K, Schillinger D. Health care discrimination, processes of care, and diabetes patients' health status. Patient Educ Couns. 2006;60(1):41-48. [PubMed: 16332469]

35.

Peek ME, Wagner J, Tang H, Baker DC, Chin MH. Self-reported racial/ethnic discrimination in healthcare and diabetes outcomes. Med Care. 2011;49(7):618-625. [PMC free article: PMC3339627] [PubMed: 21478770]

36.

Ryan AM, Gee GC, Griffith D. The effects of perceived discrimination on diabetes management. J Health Care Poor Underserved. 2008;19(1):149-163. [PubMed: 18263991]

37.

Lee C, Ayers S, Korenenfeld J. The association between perceived provider discrimination, health care utilization, and health status in racial and ethnic minorities. Ethn Dis. 2009;19(3):330-337. [PMC free article: PMC2750098] [PubMed: 19769017]

38.

Casagrande SS, Gary TL, LaVeist TA, Gaskin DJ, Cooper LA. Perceived discrimination and adherence to medical care in a racially integrated community. J Gen Intern Med. 2007;22(3):389-395. [PMC free article: PMC1824749] [PubMed: 17356974]

39.

Gaskin DJ, Price A, Brandon DT, LaVeist TA. Segregation and disparities in health services use. Med Care Res Rev. 2009;66(5):578-589. [PMC free article: PMC3099538] [PubMed: 19460811]

40.

Burgess D, Ding Y, Hargreaves M, van Ryn M, Phelan S. The association between perceived discrimination and underutilization of needed medical and mental health care in a multi-ethnic community sample. J Health Care Poor Underserved. 2008;19(3):894-911. [PubMed: 18677077]

41.

Sonik RA, Creedon TB, Progovac AM, et al. Depression treatment preferences by race/ethnicity and gender and associations between past healthcare discrimination experiences and present preferences in a nationally representative sample. Soc Sci Med. 2020;253:112939. [PubMed: 32276182]

42.

Delman J, Progovac AM, Flomenhoft T, Delman D, Chambers V, Cook BL. Barriers and facilitators to community-based participatory mental health care research for racial and ethnic minorities. Health Aff (Millwood). 2019;38(3):391-398. [PubMed: 30830821]

43.

Kroenke K, Spitzer RL, Williams JBW. The PHQ-9. J Gen Intern Med. 2001;16(9):606-613. [PMC free article: PMC1495268] [PubMed: 11556941]

44.

Schneider A, Pankow J, Heiss G, Selvin E. Validity and reliability of self-reported diabetes in the Atherosclerosis Risk in Communities Study. Am J Epidemiol. 2012;176(8):738-743. [PMC free article: PMC3571247] [PubMed: 23013620]

45.

Pratt L, Brody D. Depression and obesity in the US adult household population, 2005-2010. NCHS Data Brief. 2014;(167):1-8. [PubMed: 25321386]

46.

Centers for Disease Control and Prevention. National Diabetes Statistics Report 2020: Estimates of Diabetes and Its Burden in the United States. Centers for Disease Control and Prevention, US Department of Health and Human Services; 2020. Accessed November 6, 2020. https://www​.cdc.gov/diabetes​/pdfs/data/statistics​/national-diabetes-statistics-report.pdf

47.

Thomas R. Research quality: fast and furious … or much ado about nothing? Sub-optimal respondent behavior and data quality. J Advert Res. 2014;54(1):17-31.

48.

Louviere J, Hensher D, Swait J. Stated Choice Methods: Analysis and Applications. Cambridge University Press; 2000.

49.

Ryan M, Farrar S. Using conjoint analysis to elicit preferences for health care. BMJ. 2000;320(7248):1530-1533. [PMC free article: PMC1118112] [PubMed: 10834905]

50.

Bridges J, Kinter E, Kidane L, Heinzer R, McCormick C. Things are looking up since we started listening to patients: trends in the application of conjoint analysis in health 1982-2007. Patient. 2008;1(4):273-282. [PubMed: 22272995]

51.

Groenewoud S, Van Exel N, Bobinac A, Berg M, Huijsman R, Stolk E. What influences patients' decisions when choosing a health care provider? Measuring preferences of patients with knee arthrosis, chronic depression, or Alzheimer's disease, using discrete choice experiments. Health Serv Res. 2015;50(6):1941-1972. [PMC free article: PMC4693843] [PubMed: 26768957]

52.

Zimmerman T, Clouth J, Elosge M, et al. Patient preferences for outcomes of depression treatment in Germany: a choice-based conjoint analysis study. J Affect Disord. 2013;148(2-3):210-219. [PubMed: 23290792]

53.

Veldwijk J, Lambooij M, De Bekker-Grob EW, Smit H, de Wit G. The effect of including an opt-out option in discrete choice experiments. PLoS One. 2014;9(11):e111805. doi:10.1371/journal.pone.0111805 [PMC free article: PMC4218820] [PubMed: 25365169] [CrossRef]

54.

Wheeler B. Package ‘AlgDesign.’ CRAN R Project; 2019. Accessed November 6, 2020. https://cran​.r-project​.org/web/packages/AlgDesign/AlgDesign​.pdf

55.

Chrzan K, Orme B. Sawtooth Software; 2000. Accessed February 1, 2021. https:​//sawtoothsoftware.com/

56.

Delman J, Progovac AM, Flomenhoft T, Delman D, Chambers V, Cook BL. Barriers and facilitators to community-based participatory mental health care research for racial and ethnic minorities. Health Aff (Millwood). 2019;38(3):391-398. [PubMed: 30830821]

57.

Trivedi AN, Ayanian JZ. Perceived discrimination and use of preventive health services. J Gen Intern Med. 2006;21(6):553-558. [PMC free article: PMC1924636] [PubMed: 16808735]

58.

Williams DR, Yan Y, Jackson JS, Anderson NB. Racial differences in physical and mental health: socio-economic status, stress and discrimination. J Health Psychol. 1997;2(3):335-351. [PubMed: 22013026]

59.

Golden SD, Earp JAL. Social ecological approaches to individuals and their contexts: twenty years of health education & behavior health promotion interventions. Health Educ Behav. 2012;39(3):364-372. [PubMed: 22267868]

60.

Lieber E, Weisner T, Taylor J. Dedoose. Sociocultural Research Consultants; 2011. Accessed February 1, 2021. https://www​.dedoose.com/

61.

Hawkins J, Watkins DC, Kieffer E, et al. An exploratory study of the impact of gender on health behavior among African American and Latino men with type 2 diabetes. Am J Mens Health. 2017;11(2):344-356. [PMC free article: PMC5675282] [PubMed: 27923970]

62.

Courtenay W. Constructions of masculinity and their influence on men's well-being: a theory of gender and health. Soc Sci Med. 2000;50(10):1385-1401. [PubMed: 10741575]

63.

Hammond WP, Matthews D, Corbie-Smith G. Psychosocial factors associated with routine health examination scheduling and receipt among African American men. J Natl Med Assoc. 2010;102(4):276-289. [PMC free article: PMC2865157] [PubMed: 20437735]

64.

Williams D. The health of men: structured inequalities and opportunities. Am J Public Health. 2008;98(Suppl 9):S150-S157. [PMC free article: PMC2518607] [PubMed: 18687602]

65.

Matthews D, Hammond WP, Cole-Lewis Y, Nuru-Jeter A, Melvin T. Racial discrimination and depressive symptoms among African-American men: the mediating and moderating roles of masculine self-reliance and John Henryism. Psychol Men Masc. 2013;14(1):35-46. [PMC free article: PMC6197817] [PubMed: 30364828]

66.

Powell W, Adams L, Cole-Lewis Y, Agyemang A, Upton R. Masculinity and race-related factors as barriers to health help-seeking among African American men. Behav Med. 2016;42(3):150-163. [PMC free article: PMC4979354] [PubMed: 27337619]

67.

Ward E, Mengesha M. Depression in African American men: a review of what we know and where we need to go from here. Am J Orthopsychiatry. 2013;83(2 Pt 3):386-397. [PMC free article: PMC4215700] [PubMed: 23889029]

68.

Hankerson S, Weissman M. Church-based health programs for mental disorders among African Americans: a review. Psychiatr Serv. 2012;63(3):243-249. [PMC free article: PMC3952019] [PubMed: 22388529]

69.

Luque J, Ross L, Gwede C. Qualitative systematic review of barber-administered health education, promotion, screening and outreach programs in African-American communities. J Community Health. 2014;39(1):181-190. [PMC free article: PMC3947222] [PubMed: 23913106]

70.

Doornbos MM, Zandee GL, DeGroot J, Warpinski M. Desired mental health resources for urban, ethnically diverse, impoverished women struggling with anxiety and depression. Qual Health Res. 2013;23(1):78-92. [PubMed: 23166153]

71.

Progovac AM, Cortés DE, Chambers V, et al. Understanding the role of past health care discrimination in help-seeking and shared decision-making for depression treatment preferences. Qual Health Res. 2020;30(12):1833-1850. doi:10.1177/1049732320937663 [PMC free article: PMC10797602] [PubMed: 32713258] [CrossRef]

Related Publications

1. Delman J, Progovac A, Flomenhoft T, Delman D, Chambers V, Cook BL. Barriers and facilitators to community-based participatory mental health care research for racial and ethnic minorities. Health Aff (Millwood). 2019;38(3):391-398. [PubMed: 30830821]
2. Sonik RA, Creedon TB, Progovac AM, et al. Depression treatment preferences by race/ethnicity and gender and associations between past healthcare discrimination experiences and present preferences in a nationally representative sample. Soc Sci Med. 2020;253:112939. doi:10.1016/j.socscimed.2020.112939 [PubMed: 32276182] [CrossRef]
3. ^*Progovac AM, ^*Cortés DE, Chambers V, et al. Understanding the role of past health care discrimination in help-seeking and shared decision-making for depression treatment preferences. Qual Health Res. 2020;30(12):1833-1850. doi:10.1177/1049732320937663 [PMC free article: PMC10797602] [PubMed: 32713258] [CrossRef]
4. Adams LB, Creedon TB, Progovac AM, Rodgers CRR, Sonik R, Cook BL. Characterizing personal and intersectional discrimination in the healthcare system: a latent class analysis. (Under review, Am J Public Health)
5. Creedon TB, et al. Diabetes treatment preferences by race/ethnicity and gender and associations between past healthcare discrimination experiences and present preferences in a nationally representative sample. (In preparation) [PubMed: 32276182]
6. Flores MW, Adams LB, Tran NM, Stauffer C, Cook BL, McCormick D. The relationship between discrimination and health care quality is mediated by therapeutic relationships among individuals with type 2 diabetes. (In preparation, J Gen Intern Med)
7. Cortés DE, Progovac AM, et al. Eliciting patient treatment preferences and past experiences of discrimination: a qualitative study of primary care providers and leadership. (In preparation)

Acknowledgment

Research reported in this report was funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (#ME-1507-31469). Further information available at: https://www.pcori.org/research-results/2016/comparing-preferences-depression-and-diabetes-treatment-among-adults-different

Appendices

Suggested citation:

Cook BL, Progovac AM, Cortés DE, et al. (2021). Comparing Preferences for Depression and Diabetes Treatment among Adults of Different Racial and Ethnic Groups Who Reported Discrimination in Health Care. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/01.2021.ME.150731469

Disclaimer

The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.