Racial differences in trust and regular source of patient care and the implications for prostate cancer screening use

Abstract

BACKGROUND:

Nonmedical factors may modify the biological risk of prostate cancer (PCa) and contribute to the differential use of early detection; curative care; and, ultimately, greater racial disparities in PCa mortality. In this study, the authors examined patients' usual source of care, continuity of care, and mistrust of physicians and their association with racial differences in PCa screening.

METHODS:

Study nurses conducted in-home interviews of 1031 African-American men and Caucasian-American men aged ≥50 years in North Carolina and Louisiana within weeks of their PCa diagnosis. Medical records were abstracted, and the data were used to conduct bivariate and multivariate analyses.

RESULTS:

Compared with African Americans, Caucasian Americans exhibited higher physician trust scores and a greater likelihood of reporting a physician office as their usual source of care, seeing the same physician at regular medical encounters, and historically using any PCa screening. Seeing the same physician for regular care was associated with greater trust and screening use. Men who reported their usual source of care as a physician office, hospital clinic, or Veterans Administration facility were more likely to report prior PCa screening than other men. In multivariate regression analysis, seeing the same provider remained associated with prior screening use, whereas both race and trust lost their association with prior screening.

CONCLUSIONS:

The current results indicated that systems factors, including those that differ among different sources of care and those associated with the continuity of care, may provide tangible targets to address disparities in the use of PCa early detection, may attenuate racial differences in PCa screening use, and may contribute to reduced racial disparities in PCa mortality. Cancer 2009. Published 2009 by the American Cancer Society.

Prostate cancer (PCa) presents 1 of the greatest cancer burdens for men and accounts for the highest incidence and the second highest number of deaths from cancer among men in the United States.1 The PCa incidence rate among African Americans (AAs) is 55% greater than the rate among Caucasian Americans (CAs), and the AA mortality rate is an astounding 2.5 times that of CAs.1 Genetic and dietary factors have been identified that explain a portion of the excess burden experienced by AAs,2-5 yet risk factors that are both of substantial magnitude and amenable to preventive intervention have not been identified to date.6

In addition to biological risk factors, various other factors, including mistrust of physicians, lack of a usual source of care, and lack of continuity of care have been associated with reduced use of preventive and curative services, delayed care, and disparities in cancer and other health outcomes.7-12 AA patients report lower levels of trust in healthcare providers than CA patients, and both personal healthcare experiences and past discrimination have been implicated.13-16 It has been demonstrated that racial discrimination reduces the continuity of care and increases mistrust of providers,14, 15, 17-19 which may reduce AAs' use of recommended treatments, hinder the provision or acceptance of physician advice to undergo screening, or reduce discussion of the risks and benefits of PCa screening and early detection.20-24 AAs' lower trust, lower likelihood of having any regular source of care or using private physician offices for regular care, and lower likelihood of having an ongoing relationship with the same provider may contribute to delayed diagnosis and greater disease severity among AAs.25-32 Although PCa screening remains controversial, these 3 factors—mistrust, regular source of care, and continuity of care—are potential points of intervention and, thus, merit examination to enhance our understanding of racial differences in PCa screening and outcomes.

Grounded in a framework described by the Institute of Medicine (IOM),7 this report examines the association between mistrust of physicians, patients' usual source of care, continuity of care, and differential use of PCa screening in a large cohort of CA and AA men with recently diagnosed PCa. Previous work proposed that AAs without a regular source of care may be less likely to know their providers well enough to develop trust, thus reducing their likelihood of discussing and subsequently receiving PCa screening.33 This work broadens the examination to include systems and process factors that may be associated with both PCa screening and trust with the objectives of informing our understanding of contributors to racial disparities in PCa outcomes and improving public health and policy interventions to reduce those disparities.

MATERIALS AND METHODS

Patient Population and Sample

Methodology of the North Carolina-Louisiana Prostate Cancer Project (PCaP) was described in detail previously.33 Briefly, PCaP is a population-based study of individuals who were identified shortly after a PCa diagnosis. Study enrollment began in both states in September 2004. The original intent was to contemporaneously enroll equally from each state; however, study enrollment was stopped in Louisiana because of Hurricane Katrina (it resumed 2 years later), which substantially disrupted the healthcare systems, practices of providers that interfaced with the prospective study population, and the local population itself, causing substantial differences between the pre-Katrina and post-Katrina populations. Thus, post-Katrina Louisiana participants were not included in the current analysis. The extent and implications of this disruption are under ongoing examination, and individuals who enrolled after Katrina will be included carefully in future analyses.

Eligible individuals were identified in North Carolina through the Rapid Case Ascertainment Core Facility, a collaborative program of the University of North Carolina-Lineberger Comprehensive Cancer Registry and the North Carolina Central Cancer Registry, and, in Louisiana, through the Louisiana Tumor Registry in the Louisiana State University Health Science Center School of Public Health. Both programs allow case ascertainment and enrollment that is much faster and more complete than most traditional registries. Prospective participants typically were identified within weeks of their diagnosis, and after randomization and physician notification, study participants were enrolled. An appointment was scheduled with a study nurse (in-home or at a location of their choosing) to obtain consent, collect biospecimens, and conduct a 749-question structured survey. The visits averaged slightly more than 3 hours, including 2 breaks.

Because organizations that offer screening recommendations generally recommend screening begin at age 50 years,34 the analytic sample was limited to men aged ≥50 years (maximum age, 79 years). Our analytic sample size was 1031 men (503 AA men, 528 CA men), including men from North Carolina for whom enrollment, interviews, and primary data collection were complete through May 2007and men from Louisiana who had enrolled before Hurricane Katrina (August 29, 2005).

Analytic Framework

The IOM provides a framework for examining contributors to health disparities. In addition to patient-level factors, the IOM describes healthcare systems factors, process factors, and the dynamic interplay of patient and provider attitudes and behaviors.7, 8, 35 Systems factors include the availability of healthcare facilities, the services offered at those facilities, the systems in place to trigger appropriate use of those services, and clinician time pressures or encounter characteristics that may impede their ability to fully address patient needs.7 Process factors include provider bias, erroneous stereotypes, or lack of understanding of minorities, all of which can contribute to decision making based on incomplete or inaccurate information. All of these factors may influence patient trust, health behaviors, and receptivity toward seeking or using particular healthcare services,7 including PCa screening.

Measures

Two outcome variables—trust, and prediagnostic PCa screening use—were examined using sequential analytic models. Trust is a 100-point index measure of patient-physician trust based on a 12-item patient survey instrument that was adapted from the instruments described by Safran et al36 and LaVeist et al.17 Factor analysis of the trust instrument demonstrated good internal consistency (0.77) and loaded on 2 factors, 1 characterized as “concern with physicians hiding mistakes and bias in care delivery,” and the other with “a theme of candor, disclosure, and personal concern.” Prediagnostic PCa screening is a binary measure that summarizes patient reported prediagnostic prostate-specific antigen (PSA) and digital rectal examination (DRE) screening. Men who reported the use of either prediagnostic PSA, DRE, or both before the interaction that led to the diagnosis of PCa were coded as having had PCa screening. This “either-or-both” measure was used for 2 reasons: 1) Both PSA and DRE are prescribed by the groups that offer guidelines regarding PCa screening and early detection, and 2) excluding either PSA or DRE may bias associations of interest, because use of PSA and DRE for screening/early detection differs by race. Trust was examined as the dependent variable in Model 1 and as an independent variable in Model 2.

Two independent variables capture systems factors: Seeing the same care provider at each visit, a measure of prediagnosis care patterns and continuity of care, was based on the survey question, “Did you see a particular physician, nurse, or other medical person (at your usual source of care), or did you see a different person at each visit?” Similarly, patient's usual source of care was examined based on the 8 response-item interview question, “Before you were diagnosed with PCa, what kind of place did you usually go to when you had a medical problem?”

Several patient health, social, and economic characteristics that have been associated previously with care-seeking behaviors, support for care-seeking, and access to care were examined (Table 1). The Charlson index of historical health issues was constructed from a comorbidity questionnaire.37, 38 By using the 12-item short form (SF-12 Health Survey), version 2,39 current physical and mental health and well being were characterized, because these items may influence current perceptions and survey response. The Rapid Estimate of Adult Literacy in Medicine (REALM) questionnaire was used to assess health literacy.40, 41 Patient-physician communication was assessed using a 5-item instrument and, along with the trust measure, to describe process and patient-provider interaction characteristics. A binary measure of state of residence (North Carolina or Louisiana) was included to control for systematic differences by state.

Table 1. Characteristics of Study Sample Population Overall, and by Race

Characteristic	Overall		Caucasian American		African American		P*
	No.	%	No.	%	No.	%
SD indicates standard deviation; PCa, prostate cancer; PSA, prostate-specific antigen; DRE, digital rectal examination; REALM, Rapid Estimate of Adult Literacy in Medicine; SF-12, the SF-12 Health Survey. * T tests or chi-square tests were used to determine differences between African-American men and Caucasian men.
Overall sample size	1031		528	51.2	503	48.8
Prediagnosis PCa screening history
PSA screening	694	68.4	425	81.2	269	54.7	<.001
DRE screening	856	83.7	469	89.2	387	77.9	<.001
Either screening	894	87.8	491	93.9	403	81.4	<.001
Trust score	60.67	11.85	62.73	11.27	58.48	12.06	<.001
Sees the same clinician each time	867	84.1	472	89.4	395	78.5	<.001
Usual source of care							<.001
Physician office	798	77.5	467	88.5	331	65.9
Hospital clinic	59	5.7	17	3.2	42	8.4
Veterans Administration	93	9.0	20	3.8	73	14.5
Public health/community health center	29	2.8	7	1.3	22	4.4
ER/urgent care/none	51	5.0	17	3.2	34	6.8
Age in years, mean±SD	63.65 ± 7.48		64.89 ± 7.48		62.35 ± 7.27		<.001
Income							<.001
<$30,000	344	33.4	112	21.2	232	46.1
$30-$59,000	298	28.9	149	28.2	149	29.6
>$59,000	326	31.6	239	45.3	87	17.3
Missing data	63	6.1	28	5.3	35	7
No. with health insurance	899	87.2	502	95.1	397	78.9	<.001
Education level							<.001
>High school	550	53.4	362	68.6	188	37.4
High school	259	25.1	111	21	148	29.4
<High school	222	21.5	55	10.4	167	33.2
REALM Health Literacy							<.001
≥High school	611	59.3	430	81.4	181	36
Seventh-eighth grades	160	15.5	53	10	107	21.3
Fourth-sixth grades	146	14.2	30	5.7	116	23.1
≤Third grade	114	11.1	15	2.8	99	19.7
Communication score, mean±SD	87.29 ± 15.19		88.65 ± 15.08		85.81 ± 15.18		.001
Marital status							<.001
Married/living as married	769	74.6	433	82	336	66.8
Separated/divorced	142	13.8	47	8.9	95	18.9
Widowed	66	6.4	31	5.9	35	7
Never married	54	5.2	17	3.2	37	7.4
No. with family history of PCa	232	22.5	110	20.8	122	24.3	.189
SF-12 score, mean±SD
Physical	44.13 ± 11.28		44.95 ± 11.05		43.27 ± 11.46		.009
Mental	52.11 ± 10.94		52.95 ± 9.87		51.22 ± 11.90		.006
Charlson Comorbidity Index
Charlson score, mean±SD	0.90 ± 1.22		0.83 ± 1.10		0.98 ± 1.32		.018
0	516	50.1	274	51.9	242	48.1	.379
1	274	26.6	139	26.3	135	26.8
≥2	241	23.4	115	21.8	126	25.1
Gleason score, mean±SD	6.51 ± 0.98		6.46 ± 0.93		6.57 ± 1.02		.044
Stage at diagnosis							.240
I	536	58	264	56.5	272	59.4
II	370	40	197	42.2	173	37.8
III	12	1.3	4	0.9	8	1.8
IV	7	0.8	2	0.4	5	1.1
State of residence							.020
NC	822	79.7	436	82.6	386	76.7
La	209	20.3	92	17.4	117	23.3

Analysis

Student t tests, Fisher exact tests, and chi-square tests were used to test differences in mean values and categorical distributions of multicategory binary variables, as appropriate. The primary analysis consisted of 3 stages. First, linear regression was used to analyze the association between covariates and the dependent variable trust. Second, logistic regression was used to analyze the association between covariates, including trust, and the dependent variable, PCa screening use. White tests indicating the presence of heteroskedasticity prompted the use of robust standard errors (Huber/White sandwich estimator of variance) in both models.42 Finally, differences in a history of previous screening in association with continuity of care and usual source of care (care site) were examined independently through binomial linear regression models with and without adjustment or stratification by race.

Proper functional form of variables in the analytic models was examined, and continuous variables with non-normal distributions or nonlinear association with the dependent variables were reshaped into multicategory binary variables or were coupled with their quadratic term. The results of Wooldridge tests43 of functional form of the trust variable indicated that deviating from the natural form did not seem to add value; accordingly, the variable was used in its natural form. Multicategory response measures (eg, income, education, health literacy, marital status, type of health insurance) were examined with regard to explanatory power, magnitude, and direction of covariate associations and subsequently were collapsed into fewer categories to preserve degrees of freedom and statistical power in the analytic models. Health insurance was examined independently and in combination with age to assess differences among different types of insurance and structural differences between men aged <65 years and men aged ≥65 years who might have been eligible for Medicare. Deviating from a binary measure (any insurance vs no insurance) did not improve overall model fit.

RESULTS

CAs had higher physician trust scores than AAs (62.7 vs 58.5) and had greater past use of PCa screening of any kind (93.9% vs 81.4%) (Table 1). Most participants saw the same provider at each medical encounter, although CAs did so more often than AAs (89.4% vs 78.5%), and CAs were more likely than AAs to report a physician office or group practice as their usual source of care (88.5% vs 65.9%). At diagnosis, CAs were older, wealthier, more likely ever to have been married, and more likely to have health insurance. Respondents were well educated, and most had completed or gone beyond high school, although CAs had received more formal education, had higher REALM health literacy scores, and reported better physician communication than AAs. AAs had lower current physical, mental health, and global well being scores than CAs, but the measured differences were small, and comorbidity and family history of PCa were similar. Stage at diagnosis was similar between races, but mean Gleason scores were slightly higher for AAs than for CAs. The link between screening and tumor characteristics was not examined extensively; however, men who reported a prior history of screening were more likely than other men to be diagnosed with local or regional stage PCa (98.5% vs 94.3% stage I or II, P < .01) and with lower grade PCa (88.3% vs 81.6% with Gleason scores <8; P < .05). Participants were similar between states, and moderate differences were observed among a few measures. In bivariate analysis, North Carolina participants reported higher income and were more likely to have insurance (88.6% vs 82.1%), more formal education (56.8% vs 39.1% had an education level greater than high school graduate), and a history of PCa screening (89.7% vs 80.9%). At the same time, North Carolina participants had slightly, although statistically significantly, lower scores than Louisiana participants for communication (86.9 vs 89.6), health literacy (0.71 vs 1.0), and trust (60.2 vs 62.9).

In regression analyses, overall mean levels of trust varied in association with patient characteristics; however, differences in mean scores were small. Relative to other covariates, differences in trust by race were larger, and AAs expressed lower trust than CAs (Table 2). Communication and greater self-reported mental health and well being had a positive association with trust (a 2.3-unit increase in trust with every 10-unit increase in communication), whereas men who were high school graduates or beyond expressed lower trust than less educated men (−1.90), and the same was true for men from North Carolina relative to men from Louisiana (−2.53).

Table 2. Association of Covariates With Patient-Reported Trust of Physician (Linear Regression Model 1) and Association of Covariates With Prediagnostic North Carolina-Louisiana Prostate Cancer Project Screening Use (Logistic Regression Model 2)

Variable	Model 1: Linear Regression, Trust on Covariates		Model 2: Logistic Regression, PCa Screening on Trust and Covariates
	Coefficient	P	OR	95% CI
PCa indicates prostate cancer; OR, odds ratio; 95% CI, 95% confidence interval; SF-12, the SF-12 Health Survey; MSE, mean square error.
Trust	Dependent variable in this model		1.000	0.979-1.021
Sees same clinician	1.738	.100	2.397	1.468-3.916
African American race	−3.540	<.001	1.011	0.579-1.766
Communication	0.229	<.001	1.011	0.997-1.025
Ever married	0.521	.705	2.494	1.140-5.459
Age	0.028	.613	1.015	0.977-1.056
Income, US$
<29k	−0.221	.834	0.238	0.122-0.465
30-59k	−1.508	.093	0.972	0.411-2.295
≥$60k	Reference		Reference
Insurance	0.664	.582	1.972	1.165-3.338
Family history of PCa	0.055	.949	1.330	0.736-2.404
Education >high school graduate	−1.902	.021	2.264	1.249-4.105
Health literacy >6th grade	−0.139	.890	2.116	1.263-3.545
SF-12
Physical	0.053	.112	0.998	0.977-1.018
Mental	0.107	.003	1.012	0.991-1.033
Charlson score
0	0.262	.772	0.617	0.338-1.128
1	0.407	.695	0.606	0.309-1.189
>1	Reference		Reference
State of residence is NC	−2.525	.005	1.389	0.829-2.328
Constant	33.645	<.001	—	—
No. of observations	979		967
F(16, 962)/Wald chi-square(18)	10.80		126.42
P>F/P>chi-square	.0000		.0000
R^2/pseudo R^2	.1509		.2633
Root MSE/log-pseudo-likelihood	10.996		−250.89

Results from the logistic regression analysis of prior PCa screening on trust and covariates (Table 2) suggested that trust was not associated directly with PCa screening; however, seeing the same provider (odds ratio [OR], 2.40; 95% confidence interval [CI], 1.47-3.92), ever having been married (OR, 2.49; 95% CI, 1.14-5.46), and having any form of health insurance (OR, 1.97; 95% CI, 1.17-3.34) all were associated with prior PCa screening. Low income (vs high income: OR, 0.24; 95% CI, 0.12-0.47) was associated negatively with screening, whereas men who had a high school or higher level of education were more than twice as likely to report screening as less educated men (OR, 2.26; 95% CI, 1.25-4.11), and the same was true for men with health literacy scores above the sixth grade (relative to men with lower health literacy scores: OR, 2.12; 95% CI, 1.26-3.55).

When examining trust, consistency of provider, and usual source of care (Table 3), mean trust was slightly greater for men who saw the same provider at each visit relative to men who did not (mean trust scores, 61.04 vs 58.75). In both circumstances, CAs had slightly higher mean trust scores than AAs (saw the same provider: 60.96 vs 57.58, respectively; did not see the same provider: 62.94 vs 58.72, respectively). Mean trust scores were lower for men who reported an emergency room, urgent care center, or “none” as their regular care site (57.90) and for men who used a Veterans Administration (VA) facility (58.87) than for men whose usual care site was a physician practice (60.91), a hospital clinic (62.17), or a public health/community clinic (61.90). Differences in mean trust scores according to the source of usual care were comparable between CAs and AAs.

Table 3. Trust by Consistency of Same Provider and Usual Source of Care

Participant Trust of Physician	Overall		Caucasian Americans		African Americans		P*
	Mean±SD Trust Score	No.	Mean±SD Trust Score	No.	Mean±SD Trust Score	No.
SD indicates standard deviation; ER, emergency room. * The Student t test was used to determine the difference in trust between races.
See same provider each visit?
No	58.75 ± 12.39	162	60.96 ± 11.67	56	57.58 ± 12.66	106	<.05
Yes	61.04 ± 11.71	856	62.94 ± 11.21	469	58.72 ± 11.90	387	<.01
Overall	60.67 ± 11.85	1018	62.73 ± 11.27	525	58.48 ± 12.06	493	<.01
Usual source of care
Physician office	60.91 ± 11.68	789	62.70 ± 11.24	464	58.36 ± 11.84	325	<.01
Hospital clinic	62.17 ± 11.99	59	67.79 ± 12.38	17	59.89 ± 11.18	42	<.01
Veterans Administration	58.87 ± 13.20	92	60.73 ± 10.90	20	58.35 ± 13.80	72
Public health/community health center	61.90 ± 11.47	27	59.17 ± 11.66	7	62.85 ± 11.54	20
ER/urgent care/none	57.90 ± 11.65	50	62.30 ± 10.89	17	55.64 ± 11.53	33	<.05
Overall	60.68 ± 11.85	1017	62.73 ± 11.27	525	58.49 ± 12.07	492	<.01

Men who usually saw the same provider reported greater use of PCa screening compared with men who did not usually see the same provider both overall (91% vs 71.0%, respectively; prevalence difference [PD], −20; 95% CI, −27.3, −12.8), and by race (CAs: PD, −19.1; 95% CI, −30.3, −7.9; AAs: PD −17.2; 95% CI −26.7, −7.6), and no statistically significant difference was observed in the association between seeing the same provider and PCa screening according to race (Tables 4 and 5). Prior PCa screening use was more likely among men whose regular source of care was 1 of 3 sites—a physician office, hospital clinic, or the VA health system (90.9%; 95% CI, 89.0-92.7)—compared with men who received care at public health or community clinics, received care at an emergency room or urgent care center, or reported no usual source of care (51.9%; 95% CI, 40.8-63.1). Among CAs, 95% of men at the 3 sites reported having had PCa screening compared with 63% at the other sites (prevalence difference, 34%; OR, 0.66; 95% CI, 0.48-0.89). The relative difference was somewhat greater among AA men, with 86% reporting having had PCa screening at the 3 sites compared with 47% at the other sites (prevalence difference, 45%; OR, 0.55; 95% CI, 0.41-0.73). The difference between CAs (34%) and AAs (45%) was not statistically significant for this study population (Mantel-Haenszel chi-square test; p = .41).

Table 4. History of Prostate Cancer Screening by Consistency of Provider Seen*

Sees Same Provider Each Visit	Overall				Caucasian Americans				African Americans
	No. Screened	Total No.	% Screened	95% CI, %	No. Screened	Total No.	% Screened	95% CI, %	No. Screened	Total No.	% Screened	95% CI, %
95% CI indicates 95% confidence interval; PD, prevalence difference. * P = .79 for a difference in the association between no regular provider and screening by race.
Same Provider
Yes	779	856	91	89.1-92.9	448	467	95.9	94.1-97.7	331	389	85.1%	81.6-88.6
No	115	162	71	64-78	43	56	76.8	65.7-87.8	72	106	67.9	59-76.8
Total	894	1018			491	523			403	495
PD, %
Crude			−20	−27.3, −12.8			−19.1	−30.2, −7.9			−17.2	−26.7, −7.6
Controlling for race			−18	−25.2, −10.8

Table 5. History of Prostate Cancer Screening by Usual Source of Care*

Usual Source of Care	No. Screened	Total No.	% Screened,	95% CI, %	PD (95% CI), %
					Crude	Controlling for Race
PD indicates prevalence difference; 95% CI, 95% confidence interval; ER, emergency room. * P = 64 for a difference in the association between no regular provider and screening by race.
Overall
Physician office	723	790	91.5	89.6-93.5	0
Hospital clinic	49	58	84.5	75.2-93.8	−7 (−16.6, 2.5)
Veterans Administration	82	92	89.1	82.8-95.5	−2.4 (−9, 4.3)
Public health/community health center	16	28	57.1	38.8-75.5	−34.4 (−52.8, −15.9)
ER/urgent care/none	24	49	49.0	35-63	−42.5 (−56.7, −28.4)
Total	894	1017	87.9
Consolidated source of care by race
Overall					−38.9 (−50.2, −27.6)	−36.7 (−48.1, −25.3)
Physician office/hospital/Veterans Administration	854	940	90.9	89-92.7
Other	40	77	51.9	40.8-63.1
Caucasian Americans					−32.9 (−52.3, −13.4)
Physician office/hospital/Veterans Administration	476	499	95.4	93.6-97.2
Other	15	24	62.5	43.1-81.9
African Americans					−38.5 (−52.4, −24.7)
Physician office/hospital/Veterans Administration	378	441	85.7	82.4-89
Other	25	53	47.2	33.7-60.6

DISCUSSION

To gain a better understanding of racial disparities in PCa mortality, we examined relations between PCa patients' trust in their physicians, and continuity of healthcare, regular source of healthcare, and their use of PCa screening. In our study population of patients with newly diagnosed PCa, AAs were less likely than CAs to report PCa screening before diagnosis, and men without a prior history of screening were more likely to be diagnosed with advanced-stage or high-grade PCa than men who reported a history of PCa screening. Although the efficacy of PSA screening remains controversial,34 these findings support the hypothesis that addressing differences in screening use may reduce racial disparities in PCa extent of disease and aggressiveness at diagnosis, and, by extension, that addressing differences in screening may reduce disparities in PCa mortality.44 Adjusting for patient-provider interaction, health systems, and processes sharply attenuated the association between race and PCa screening history in multivariate analysis. Previously, we reported an association between having a consistent care provider and trust in healthcare providers45; in the current study of patients with PCa, seeing a regular provider was associated more closely with prior use of PCa screening, while trust was not associated with screening in our multivariate analysis.

This analysis examined men who already were diagnosed with PCa, whose prerequisite access to the healthcare system may have required some baseline level of trust in their physicians and the healthcare community. The results indicated that, upon passing this threshold, however, other systems variables (including setting and continuity of care) predict the use of screening. Thus, the current study results reframe the question: To what degree are lower trust and lesser use of screening a function of individual patient and physician characteristics versus systems factors that may hinder trust and relationship development?

Physician biases or practical constraints may inhibit their ability to inquire beyond the primary clinical issue at hand to secondary problems, preventive and health maintenance issues, or other more complex medical decisions. An ongoing patient-provider relationship may enable the relationship to grow past the tipping point of sufficient strength to discuss these issues and overcome barriers to screening use, such as fear of the examination itself, prospective diagnosis, or potential treatment side-effects.8 On a more practical level, such a relationship simply may expedite provider-patient interaction, allowing additional time and resources to address more complex issues over time, including decisions regarding the use of preventive care, such as cancer screening or, as described by other studies, vaccinations.24

Discussions regarding screening and early detection also may be promoted by office and support systems. For example, patient records may provide reminders of prior visits and other clinical conditions and may include prompts for patient-appropriate services. Differences in the use of PCa screening have been reported according to the regular source of care. For example, prior research has demonstrated the weaknesses of emergency departments and urgent care centers for the provision of preventive care.27-30 Simultaneous analyses of regular source of care and seeing the same provider suggested that seeing the same provider may partially overcome observed differences between public and community health clinics, private practice, hospital clinics, and VA facilities to foster the use of preventive care. Men who reported not seeing the same provider also reported less PCa screening use; however, data limitations impeded a precise demonstration of the relation among the care sites.

These findings suggest the need for additional research examining the roles of provider continuity and the systems in place at the usual source of care, possibly directed toward an intervention that targets healthcare organizations. The limitations of the current study also suggest additional avenues. Having health insurance improves the use of both early detection and treatment-oriented health services.35 Characterizing prediagnosis insurance beyond our indicators “any insurance” and “no insurance” may clarify the expected association between having insurance and use of PSA/DRE before diagnosis that we observed in this study. Differences in insurance between North Carolina and Louisiana and changes in men's insurance after diagnosis also merit examination. Insurance coverage is relevant for several reasons, including the possibility that it may influence the reporting of DRE compared with PSA, which may be related directly to the type, timing, and duration of insurance coverage. The reported use of PSA (which is billable separately) may be low compared with DRE use (which is included in comprehensive physical examinations), because PSA is used less commonly in the absence of consistent financial resources. PSA use also may be under-reported, because a PSA test can be added to a panel of other blood tests without the patient knowing it; whereas DRE rarely goes unnoticed, although it conceivably could be confused with examinations for other conditions. For researchers who examine PCa screening, insurance claims and medical records may be used to augment these self-reported data, but insurance claims only capture PSA testing for the insured, a comprehensive gathering of medical records is expensive, and the documentation of nonbillable procedures such as DRE often is spotty.

Similar to our measures of insurance status, our measure of trust could be enhanced. Trust was assessed at a single point in time (postdiagnosis) and was analyzed in the context of antecedent health behavior (prediagnostic screening use). Trust measured at this point may be different and less relevant than trust before diagnosis or earlier, when men's healthcare behaviors may have formed. These findings generally support the conclusions of previous research indicating that patients who have familiarity, better communication, and longer relationships or more visits with their physicians report comparatively greater trust.15, 46-50 Moreover, the data encourage more nuanced characterization of both race and trust than as monolithic, homogeneous, and invariant. Instead, AAs, similar to CAs, are both socially vulnerable and heterogeneous, and the trust they exhibit is a dynamic function of multiple characteristics that may vary over time and by context.

The attenuation of racial differences in screening use observed in the multivariate model suggests that approaching the problem from multiple angles holds merit; however, the persistence of disparities in all strata of trust and structure of usual care reinforces the complexity of the problem. Adapting systems to assure provider continuity may facilitate patient-provider interactions that extend beyond the primary problem at hand. The adoption of additional systems to prompt and facilitate preventive care may provide additional benefit, especially among public health and community clinics, and even in emergency rooms and urgent care centers that acknowledge and accept that they are the primary care providers for some individuals.51, 52 The means of enhancing trust among the broader population remains unclear, although the current results suggest that addressing systems issues may be an important route of intervention to improve trust among those who access healthcare services. Doing so may contribute to trust and attenuate racial differences in the short term, which, in turn, may feed back into the community and contribute to a resolution of differences in trust, access, and outcomes in the long term for the broader population.

Acknowledgements

We thank Jane Schroeder, DVM, PhD, for her assistance in dataset development and interpretation. We also thank the staff, advisory committees, and the patients w ho participated in the North Carolina-Louisiana Prostate Cancer Project for their important contributions.

Conflict of Interest Disclosures

The North Carolina-Louisiana Prostate Cancer Project is performed as a collaborative study supported by the Department of Defense contract DAMD 17-03-2-0052.

Dr. Carpenter was supported by a training grant from the National Cancer Institute (R25T CA 57726).