Abstract
Background
Determining whether observed differences in healthcare can be called disparities requires persistence of differences after adjustment for relevant patient, provider and health system factors. We examined whether providing dual energy x-ray absorptiometry (DXA) test results directly to patients might reduce or eliminate racial differences in osteoporosis-related healthcare.
Design, Subjects, and Measures
We analyzed data from 3,484 White and 1,041 Black women who underwent DXA testing at two health systems participating in the Patient Activation after DXA Result Notification (PAADRN) pragmatic clinical trial (ClinicalTrials.gov NCT-01507662) between February 2012 and August 2014. We examined seven outcomes related to bone health at 12- and 52-weeks post-DXA: (1) whether the patient correctly identified their DXA baseline results; (2) whether the patient was on guideline-concordant osteoporosis pharmacotherapy; (3) osteoporosis-related satisfaction; (4) osteoporosis knowledge; (5 and 6) osteoporosis self-efficacy for exercise and for diet; and, (7) patient activation. We examined whether unadjusted differences in outcomes between Whites and Blacks persisted after adjusting for patient, provider and health system factors.
Results
Mean age was 66.5 years and 29% were Black. At baseline Black women had less education, poorer health status, and were less likely to report a history of osteoporosis (p < 0.001 for all). In unadjusted analyses Black women were less likely to correctly identify their actual DXA results, more likely to be on guideline concordant therapy, and had similar patient activation. After adjustment for patient demographics, baseline health status and other factors, Black women were still less likely to know their actual DXA result and less likely to be on guideline-concordant therapy, but Black women had greater patient activation.
Conclusions
Adjustment for patient and provider level factors can change how racial differences are viewed, unmasking new disparities and providing explanations for others.
Trial Registration
ClinicalTrials.gov Identifier: NCT01507662
Keywords: Osteoporosis, Dual Energy X-ray Absorptiometry, Randomized Controlled Trials, Risk Communication
Introduction
Two million osteoporotic fractures occur each year in the United States (U.S.) as a result of low bone mineral density (BMD) and pose a major public health problem.1 Of these, 310,000 are hip fractures.2,3 Hip fractures often occur in older women who have osteoporosis (diagnosed or not) and who fall.4,5 While White women have higher rates of hip fracture, Black women have poorer outcomes including higher morbidity and mortality.6–8 The primary preventive strategy for reducing fractures in the U.S. is screening for osteoporosis using dual-energy x-ray absorptiometry (DXA).9
There are substantial racial differences in DXA screening and osteoporosis diagnosis and treatment rates.6 For example, Curtis et al. report that among traditional Medicare beneficiaries, 33% of White vs. 5% of Black women had undergone osteoporosis screening. Cheng reported that among traditional Medicare beneficiaries with fractures, osteoporosis was diagnosed nearly twice-as-often for White women compared to Black women across all age groups.10 Hamrick reported that while 80% of White women received pharmacotherapy after osteoporosis diagnoses, only 68% of Black women did.11
What is not clear is whether these racial differences constitute disparities. In a groundbreaking 2003 U.S. Institute of Medicine (IOM) report, Smedley et al. defined racial disparities as “racial or ethnic differences in the quality of healthcare that are not due to access-related factors or clinical needs, preferences, and appropriateness of intervention.”12 That is, racial differences may be due to patient, provider, and/or institutional factors, and only constitute disparities if they exist after adjustment for these factors. Patient factors include the social determinants of health and health behaviors, provider factors include practice patterns and explicit and implicit biases, and institutional factors include access to and the organization of healthcare. The dearth of detailed information about these patient, provider, and institutional factors in administrative claims and medical records hinder efforts to determine if racial differences are true disparities.
We used data from the Patient Activation after DXA Report Notification (PAADRN) randomized controlled trial (RCT)21,22 to investigate and differentiate racial differences and disparities in bone health care among White and Black women. PAADRN data are well suited to this task for two reasons. First, PAADRN’s intervention focused on engaging patients in their bone health self-care, which has been shown in prior studies to be lower for Blacks than for Whites.13 Second, PAADRN collected an extensive battery of data about numerous aspects of bone health knowledge, self-efficacy, and behaviors.14 Accordingly, we use the PAADRN data on White and Black women to explore these issues. Building on the recommendations of the IOM report,1 we estimate four multivariable models for each of the seven bone health measures. Through successive models, we provide insight into whether racial disparities in bone health can be explained by differences in patient and provider level factors.
Methods
Design and Sample
PAADRN was a pragmatic RCT whose design has been detailed elsewhere.15,16 PAADRN enrolled 7,749 patients presenting for DXA between February 2012 and August 2014 at the University of Iowa (UI), the University of Alabama at Birmingham (UAB), and Kaiser Permanente of Georgia (KPGA). Exclusion criteria were: (1) age < 50 years; (2) prisoners or patients with overt cognitive disability; (3) patients who did not speak or read English; and, (4) patients who were deaf or lacked access to a telephone. Baseline phone or face-to-face interviews occurred up to 28-days before or 3-days after their baseline DXA, with follow-up telephone interviews occurring at 12- and 52-weeks post-DXA. Because this Brief Report focuses on racial differences and disparities, the sample excludes all patients from UI (where there were not enough Black patients for analytic purposes) and all men from UAB and KPGA (because of the uncertainty about DXA screening in men).9 Institutional Review Boards at UI, UAB, and KPGA approved the study protocol. Our full protocol is available at https://clinicaltrials.gov/ct2/show/NCT01507662.
Intervention
Intervention patients were notified of their DXA results via a tailored letter accompanied by an educational brochure.15,17,18 The letter included the clinical impression of each patient’s DXA result (normal, low BMD [osteopenia], or osteoporosis), their 10-year fracture risk, and the suggestion that the patient bring the letter to their next physician visit. The brochure explained osteoporosis, reviewed the benefits of proper calcium and vitamin D intake, exercise, fall prevention, smoking cessation, and alcohol moderation, and provided additional osteoporosis resources. Intervention materials were mailed to patients from UI 4-weeks after their baseline DXA. Usual care patients received their DXA results based on the practices of their physicians and healthcare systems.
Outcomes
We examine seven outcomes related to bone health at 12- and 52-weeks post-DXA (Table 1). Measures included: (1) whether the patient correctly identified their baseline DXA results; (2) whether the patient was on guideline-concordant osteoporosis pharmacotherapy;22 (3) osteoporosis-related satisfaction;29 (4) osteoporosis knowledge;30,31 (5 and 6) osteoporosis self-efficacy for exercise and self-efficacy for diet;26 and, (7) patient activation.31,32 All but the guideline-concordant care measure are based directly on scales of known reliability and validity, or to a shortened form (patient activation) of such a scale. Receipt of guideline-concordant pharmacotherapy occurred when patients reported taking osteoporosis pharmacotherapy when indicated, or when patients reported not taking osteoporosis pharmacotherapy when it was not indicated by their DXA results and other patient considerations according to National Osteoporosis Foundation (NOF) guidelines.22
Table 1.
Outcomes and their reliability coefficients
| Outcome | Number of Items in Scale | Range of Original Scare | Standardized Cronbach Coefficient Alpha |
|---|---|---|---|
| Correctly identified the results of their baseline DXA | 1 | 0 = No, 1 = Yes | -- |
| On guideline concordant pharmacotherapy* | 1 | 0 = No, 1 = Yes | -- |
| Osteoporosis related satisfaction 29 | 5 | 5 – 25 | 0.764 |
| Osteoporosis specific knowledge 30,31 | 10 | 0 – 10 | 0.607 |
| Osteoporosis self-efficacy exercise 26 | 10 | 0 – 100 | 0.969 |
| Osteoporosis self-efficacy diet 26 | 11 | 0 – 100 | 0.962 |
| Patient activation 31,32 | 6 | 0 – 100 | 0.719 |
Patients with osteoporosis (T-score of femoral neck, hip, or spine ≤−2.5 or FRAX ≥20 %), or with a self-reported history of low impact fracture, or with osteopenia (T-score between −1.0 and −2.5 at the femoral neck, hips, or lumbar spine) and a 10-year probability of a major osteoporosis-related fracture ≥20 % who are on osteoporosis pharmacotherapy OR patients who do not have one of the above and are not on osteoporosis pharmacotherapy. 22
Covariates
Covariates included patient, provider, and institutional characteristics (Table 2) important for differentiating racial differences and disparities according to the IOM framework.12 Patient characteristics included sociodemographics (age, sex, race, and education), comorbidities, health habits (smoking, drinking, and exercise), self-reported health, prior bone health, health literacy and numeracy, and study DXA results. Provider characteristics included sex and specialty. Institutional characteristics were limited to DXA site.
Table 2.
Baseline characteristics of White and Black women at the two PAADRN sites.
| White (N=3484) | Black (N=1041) | P-Value | |
|---|---|---|---|
| Site | |||
| UAB, number (%) | 2159 (62.0) | 620 (44.3) | <0.001 |
| KP, number (%) | 1325 (38.0) | 781 (55.7) | |
| Socio-demographics | |||
| Age, mean (SD) | 66.9 (8.11) | 66.38 (7.84) | 0.039 |
| Education | |||
| Some high school, number (%) | 96 (2.8) | 108 (7.8) | <0.001 |
| Completed high school, number (%) | 759 (21.9) | 327 (23.5) | |
| Some college, number (%) | 1157 (33.3) | 579 (41.7) | |
| Completed college, number (%) | 712 (20.5) | 220 (15.8) | |
| Graduate school, number (%) | 749 (21.6) | 155 (11.2) | |
| Comorbid Conditions | |||
| COPD, number (%) | 208 (6.0) | 79 (5.7) | 0.667 |
| Depression, number (%) | 957 (27.5) | 220 (15.7) | <0.001 |
| Breast cancer, number (%) | 613 (17.6) | 144 (10.3) | <0.001 |
| Health Habits | |||
| Current smoker, number (%) | 216 (6.2) | 119 (8.5) | 0.004 |
| Past smoker, number (%) | 1219 (35.0) | 420 (30.0) | 0.001 |
| Current alcohol user, number (%) | 872 (49.2) | 225 (31.4) | <0.001 |
| Self-reported Health Status | |||
| Excellent, number (%) | 535 (15.4) | 83 (5.9) | <0.001 |
| Very good, number (%) | 1278 (36.7) | 354 (25.3) | |
| Good, number (%) | 1170 (33.6) | 612 (43.7) | |
| Fair, number (%) | 393 (11.3) | 303 (21.7) | |
| Poor, number (%) | 106 (3.0) | 47 (3.4) | |
| Bone Health | |||
| Prior DXA, number (%) | 2903 (83.3) | 737 (52.6) | <0.001 |
| History of osteoporosis, number (%) | 898 (26.0) | 220 (15.9) | |
| History of osteoporosis treatment, number (%) | 1799 (51.6) | 319 (22.8) | |
| Health Literacy, mean (SD) | 4.36 (0.81) | 4.24 (0.93) | <0.001 |
| Health Numeracy, mean (SD) | 4.71 (0.97) | 4.12 (1.13) | <0.001 |
| Study DXA Results | |||
| Normal, number (%) | 727 (20.9) | 626 (44.7) | <0.001 |
| Low BMD, number (%) | 2002 (57.5) | 613 (43.8) | |
| Osteoporosis, number (%) | 755 (21.7) | 162 (11.6) | |
| Fracture Risk | |||
| Low, number (%) | 1181 (33.9) | 1310 (93.5) | <0.001 |
| Medium, number (%) | 1478 (42.4) | 87 (6.2) | |
| High, number (%) | 825 (23.7) | 4 (0.3) | |
| Provider Characteristics | |||
| Physician, number (%) | 3375 (96.9) | 1359 (97) | 0.811 |
| Women Provider, number (%) | 1886 (54.1) | 874 (62.4) | <0.001 |
| Provider Specialty | |||
| Family Medicine, number (%) | 512 (14.7) | 347 (24.8) | <0.001 |
| General Internal Medicine, number (%) | 1783 (51.2) | 745 (53.2) | |
| Rheumatology/Endocrinology, number (%) | 419 (12.0) | 74 (5.3) | |
| Other, number (%) | 770 (22.1) | 235 (16.8) | |
| Intervention | |||
| Intervention, number (%) | 1771 (50.8) | 716 (51.1) | 0.862 |
| Usual Care, number (%) | 1713 (49.2) | 685 (48.9) | |
| Outcomes | |||
| Correctly identified the results of their baseline DXA | |||
| 12 weeks, number (%) | 2109 (68.2) | 615 (55.2) | <0.001 |
| 52 weeks, number (%) | 1802 (63.1) | 477 (50.9) | <0.001 |
| On guideline concordant pharmacotherapy | |||
| 12 weeks, number (%) | 1934 (62.5) | 857 (76.9) | <0.001 |
| 52 weeks, number (%) | 1778 (62.3) | 713 (76.1) | <0.001 |
| Osteoporosis related satisfaction | |||
| 12 weeks, mean (SD) | 20.91 (3.77) | 20.38 (3.94) | <0.001 |
| 52 weeks, mean (SD) | 21.21 (3.64) | 20.45 (4.01) | <0.001 |
| Osteoporosis specific knowledge | |||
| 12 weeks, mean (SD) | 8.17 (1.42) | 7.10 (1.81) | <0.001 |
| 52 weeks, mean (SD) | 8.11 (1.43) | 7.09 (1.87) | <0.001 |
| Osteoporosis self-efficacy exercise | |||
| 12 weeks, mean (SD) | 7.73 (2.2) | 7.75 (2.21) | 0.790 |
| 52 weeks, mean (SD) | 7.82 (2.19) | 7.99 (2.04) | 0.040 |
| Osteoporosis self-efficacy diet | |||
| 12 weeks, mean (SD) | 8.51 (1.69) | 8.45 (1.87) | 0.368 |
| 52 weeks, mean (SD) | 8.47 (1.76) | 8.45 (1.88) | 0.812 |
| Patient activation | |||
| 12 weeks, mean (SD) | 77.34 (17.15) | 77.26 (18.48) | 0.900 |
| 52 weeks, mean (SD) | 78.13 (17.28) | 77.95 (18.25) | 0.788 |
Data analyses
Baseline characteristics of White and Black women are compared using bivariable methods. Linear mixed effects models are used to estimate four multivariable models for the seven bone health measures (after standardizing the non-binary measures). The first model only contains only race as a covariate. The second model adjusts for patient, provider, and center factors. The third model further adjusts for receipt (or not) of the mailed letter intervention. The fourth model includes the interaction of the DXA centers with the intervention and is estimated to test for potential heterogeneity of treatment effects (HTEs) based on the site characteristics, including uniform employer-based insurance and practice protocols at KPGA vs. UAB. In sensitivity analyses we restricted the sample to women ≥ 65 years old for whom osteoporosis treatment guidelines are most clearly applicable, and we used inverse probability of treatment weighting (IPTW) to adjust for the higher attrition rates among Black women. Bonferroni adjustments are used to correct for multiplicity. All p-values are 2-tailed, with those ≤ 0.025 deemed statistically significant. All statistical analyses were performed using SAS 9.4 (SAS Inc., Cary, NC).
Results
Table 2 reveals significant (all p < 0.001) baseline differences between the 3,484 White and 1,041 Black women in most domains including age, education, comorbidity, health habits, self-reported health, health literacy, DXA results, and provider characteristics. Black women are disadvantaged for many of these measures. Table 2 also reveals significant unadjusted differences between Black and White women at 12- and 52-weeks post-DXA on the osteoporosis health measures. Black women are less likely to correctly identify their DXA results and have less osteoporosis-related knowledge than White women (p < 0.001). At the same time, Black women are more likely to be on guideline-concordant pharmacotherapy and are more satisfied with their care (p < 0.001).
Table 3 contains the multivariable results for the osteoporosis health measures at 12-weeks post-DXA. Even after full adjustment (Model 3), Black women were less likely to correctly identify their DXA results (beta = −0.150, p < 0.001) and had lower osteoporosis-related knowledge (beta = −0.473, p < 0.001) than White women, suggesting health disparities.
Table 3.
Effect estimates at 12 weeks from linear mixed effect models on the seven outcomes.
| Outcome | Effect | Effect Estimate | 95% CI | P-value | |
|---|---|---|---|---|---|
| Correctly identified the results of their baseline DXA | Model 1 | Race Black vs. White | −0.145 | (−0.178, −0.111) | <0.001 |
| Model 2 | Race Black vs. White | −0.154 | (−0.194, −0.113) | <0.001 | |
| Model 3 | Race Black vs. White | −0.15 | (−0.190, −0.110) | <0.001 | |
| Intervention vs. Control | 0.118 | (0.087, 0.148) | <0.001 | ||
| Site KP vs. UAB | 0.088 | (0.052, 0.125) | <0.001 | ||
| Model 4 | Race Black vs. White | −0.151 | (−0.191, −0.111) | <0.001 | |
| Intervention vs. Control | 0.141 | (0.100, 0.181) | <0.001 | ||
| Site KP vs. UAB | 0.115 | (0.067, 0.163) | <0.001 | ||
| Site (KP) * Intervention | −0.051 | (−0.112, 0.01) | 0.099 | ||
| On guideline concordant pharmacotherapy | Model 1 | Race Black vs. White | 0.144 | (0.112, 0.176) | <0.001 |
| Model 2 | Race Black vs. White | −0.073 | (−0.111, −0.034) | <0.001 | |
| Model 3 | Race Black vs. White | −0.073 | (−0.112, −0.035) | <0.001 | |
| Intervention vs. Control | −0.016 | (−0.043, 0.01) | 0.233 | ||
| Site KP vs. UAB | −0.005 | (−0.035, 0.025) | 0.737 | ||
| Model 4 | Race Black vs. White | −0.073 | (−0.111, −0.034) | <0.001 | |
| Intervention vs. Control | −0.028 | (−0.063, 0.006) | 0.110 | ||
| Site KP vs. UAB | −0.020 | (−0.061, 0.021) | 0.335 | ||
| Site (KP) * Intervention | 0.029 | (−0.025, 0.083) | 0.289 | ||
| Osteoporosis related satisfaction | Model 1 | Race Black vs. White | −0.127 | (−0.201, −0.053) | 0.001 |
| Model 2 | Race Black vs. White | −0.053 | (−0.145, 0.039) | 0.256 | |
| Model 3 | Race Black vs. White | −0.046 | (−0.137, 0.046) | 0.328 | |
| Intervention vs. Control | 0.266 | (0.194, 0.338) | <0.001 | ||
| Site KP vs. UAB | 0.041 | (−0.05, 0.132) | 0.378 | ||
| Model 4 | Race Black vs. White | −0.046 | (−0.138, 0.045) | 0.323 | |
| Intervention vs. Control | 0.282 | (0.185, 0.38) | <0.001 | ||
| Site KP vs. UAB | 0.060 | (−0.059, 0.178) | 0.322 | ||
| Site (KP) * Intervention | −0.036 | (−0.180, 0.109) | 0.628 | ||
| Osteoporosis specific knowledge | Model 1 | Race Black vs. White | −0.656 | (−0.724, −0.588) | <0.001 |
| Model 2 | Race Black vs. White | −0.474 | (−0.556, −0.392) | <0.001 | |
| Model 3 | Race Black vs. White | −0.473 | (−0.555, −0.391) | <0.001 | |
| Intervention vs. Control | 0.021 | (−0.036, 0.079) | 0.463 | ||
| Site KP vs. UAB | 0.023 | (−0.042, 0.088) | 0.482 | ||
| Model 4 | Race Black vs. White | −0.472 | (−0.554, −0.39) | <0.001 | |
| Intervention vs. Control | −0.007 | (−0.082, 0.068) | 0.856 | ||
| Site KP vs. UAB | −0.011 | (−0.100, 0.077) | 0.801 | ||
| Site (KP) * Intervention | 0.067 | (−0.048, 0.183) | 0.253 | ||
| Osteoporosis self-efficacy exercise | Model 1 | Race Black vs. White | 0.006 | (−0.066, 0.077) | 0.873 |
| Model 2 | Race Black vs. White | 0.166 | (0.082, 0.250) | <0.001 | |
| Model 3 | Race Black vs. White | 0.166 | (0.082, 0.250) | <0.001 | |
| Intervention vs. Control | −0.010 | (−0.068, 0.048) | 0.734 | ||
| Site KP vs. UAB | 0.066 | (0.000, 0.132) | 0.050 | ||
| Model 4 | Race Black vs. White | 0.166 | (0.082, 0.250) | <0.001 | |
| Intervention vs. Control | −0.013 | (−0.090, 0.063) | 0.729 | ||
| Site KP vs. UAB | 0.062 | (−0.028, 0.152) | 0.177 | ||
| Site (KP) * Intervention | 0.008 | (−0.109, 0.125) | 0.892 | ||
| Osteoporosis self-efficacy diet | Model 1 | Race Black vs. White | −0.032 | (−0.102, 0.038) | 0.368 |
| Model 2 | Race Black vs. White | 0.082 | (−0.006, 0.17) | 0.067 | |
| Model 3 | Race Black vs. White | 0.082 | (−0.006, 0.17) | 0.069 | |
| Intervention vs. Control | −0.018 | (−0.079, 0.044) | 0.574 | ||
| Site KP vs. UAB | 0.045 | (−0.024, 0.114) | 0.203 | ||
| Model 4 | Race Black vs. White | 0.081 | (−0.007, 0.169) | 0.071 | |
| Intervention vs. Control | 0.006 | (−0.074, 0.086) | 0.877 | ||
| Site KP vs. UAB | 0.074 | (−0.020, 0.168) | 0.121 | ||
| Site (KP) * Intervention | −0.057 | (−0.180, 0.066) | 0.365 | ||
| Patient activation | Model 1 | Race Black vs. White | −0.009 | (−0.080, 0.062) | 0.796 |
| Model 2 | Race Black vs. White | 0.112 | (0.024, 0.201) | 0.013 | |
| Model 3 | Race Black vs. White | 0.114 | (0.025, 0.202) | 0.012 | |
| Intervention vs. Control | 0.05 | (−0.016, 0.116) | 0.141 | ||
| Site KP vs. UAB | 0.029 | (−0.049, 0.107) | 0.467 | ||
| Model 4 | Race Black vs. White | 0.114 | (0.026, 0.203) | 0.012 | |
| Intervention vs. Control | 0.033 | (−0.055, 0.122) | 0.463 | ||
| Site KP vs. UAB | 0.01 | (−0.094, 0.114) | 0.850 | ||
| Site (KP) * Intervention | 0.037 | (−0.095, 0.169) | 0.584 |
Model 1 includes race only (Black vs. White)
Model 2 includes race and covariates measured at baseline
Model 3 includes race, intervention, plus covariates measured at baseline
Model 4 includes race, intervention, interaction between intervention and site (KP vs. UAB), plus covariates measured at baseline
Baseline covariates include, site (KP vs. UAB), age (continuous), education, comorbidity (COPD, depression and breast cancer), self-assessed health status (using self-reported health status, continuous), heath habits (current smoker, former smoker and drinker), bone health (history of DXA, history of fracture and history of osteoporosis treatment), health literacy (continuous), health numeracy (continuous), study DXA, fracture risk, gender of provider, physician, and provider specialty.
Table 3 also reveals a suppressed disparity (i.e., a disparity that was only observed after adjustment for the covariates) and two suppressed advantages (i.e., advantages that were only observed after adjustment for the covariates) for Black women. The suppressed disparity is that prior to covariate adjustment, Black women are more likely to be on guideline-concordant pharmacotherapy (Model 1, beta = 0.144, p < 0.001). After adjusting for the covariates (especially their DXA results), however, Black women are less likely than White women to be on guideline-concordant pharmacotherapy (Model 3, beta = −0.073, p < 0.001). The two suppressed advantages for Black women reveal their greater self-efficacy for exercise (Model 3, beta = 0.166, p < 0.001) and their greater patient activation (Model 3, beta = 0.114, p = 0.012) after adjustment for the covariates, whereas the unadjusted analyses (Model 1) revealed no such advantages.
Table 4 contains the multivariable results for the osteoporosis health measures at 52-weeks post-DXA. Because these are consistent with those reported in Table 3, they are not discussed here. The sensitivity analyses (all Model 4s in Tables 3 and 4) revealed no HTEs for the intervention across the two DXA centers (all p > 0.025). The sensitivity analyses restricting our analysis to women age ≥ 65 years old and those using IPTW to adjust for differential attrition were consistent with those shown in Tables 3 and 4 (available on request).
Table 4.
Effect estimates at 52 weeks from linear mixed effect models on the eight outcomes.
| Outcome | Effect | Effect Estimate | 95% CI | P-value | |
|---|---|---|---|---|---|
| Correctly identified the results of their baseline DXA | Model 1 | Race Black vs. White | −0.128 | (−0.165, −0.091) | <0.001 |
| Model 2 | Race Black vs. White | −0.149 | (−0.193, −0.105) | <0.001 | |
| Model 3 | Race Black vs. White | −0.147 | (−0.191, −0.103) | <0.001 | |
| Intervention vs. Control | 0.065 | (0.034, 0.096) | <0.001 | ||
| Site KP vs. UAB | 0.105 | (0.068, 0.141) | <0.001 | ||
| Model 4 | Race Black vs. White | −0.147 | (−0.191, −0.103) | <0.001 | |
| Intervention vs. Control | 0.069 | (0.027, 0.11) | 0.001 | ||
| Site KP vs. UAB | 0.109 | (0.060, 0.158) | <0.001 | ||
| Site (KP) * Intervention | −0.008 | (−0.072, 0.055) | 0.798 | ||
| On guideline concordant pharmacotherapy | Model 1 | Race Black vs. White | 0.138 | (0.104, 0.173) | <0.001 |
| Model 2 | Race Black vs. White | −0.107 | (−0.148, −0.066) | <0.001 | |
| Model 3 | Race Black vs. White | −0.107 | (−0.148, −0.066) | <0.001 | |
| Intervention vs. Control | 0.003 | (−0.025, 0.031) | 0.838 | ||
| Site KP vs. UAB | −0.006 | (−0.038, 0.026) | 0.729 | ||
| Model 4 | Race Black vs. White | −0.107 | (−0.148, −0.066) | <0.001 | |
| Intervention vs. Control | 0.004 | (−0.032, 0.041) | 0.818 | ||
| Site KP vs. UAB | −0.004 | (−0.047, 0.039) | 0.859 | ||
| Site (KP) * Intervention | −0.003 | (−0.06, 0.054) | 0.909 | ||
| Osteoporosis related satisfaction | Model 1 | Race Black vs. White | −0.183 | (−0.263, −0.102) | <0.001 |
| Model 2 | Race Black vs. White | −0.085 | (−0.184, 0.014) | 0.094 | |
| Model 3 | Race Black vs. White | −0.079 | (−0.177, 0.02) | 0.117 | |
| Intervention vs. Control | 0.2 | (0.127, 0.274) | <0.001 | ||
| Site KP vs. UAB | 0.046 | (−0.043, 0.135) | 0.309 | ||
| Model 4 | Race Black vs. White | −0.078 | (−0.177, 0.021) | 0.121 | |
| Intervention vs. Control | 0.177 | (0.078, 0.276) | <0.001 | ||
| Site KP vs. UAB | 0.019 | (−0.099, 0.136) | 0.756 | ||
| Site (KP) * Intervention | 0.053 | (−0.095, 0.201) | 0.485 | ||
| Osteoporosis specific knowledge | Model 1 | Race Black vs. White | −0.625 | (−0.698, −0.553) | <0.001 |
| Model 2 | Race Black vs. White | −0.439 | (−0.527, −0.35) | <0.001 | |
| Model 3 | Race Black vs. White | −0.437 | (−0.525, −0.348) | <0.001 | |
| Intervention vs. Control | 0.058 | (−0.003, 0.12) | 0.063 | ||
| Site KP vs. UAB | 0.012 | (−0.058, 0.082) | 0.741 | ||
| Model 4 | Race Black vs. White | −0.436 | (−0.525, −0.347) | <0.001 | |
| Intervention vs. Control | 0.029 | (−0.05, 0.109) | 0.470 | ||
| Site KP vs. UAB | −0.024 | (−0.118, 0.071) | 0.623 | ||
| Site (KP) * Intervention | 0.069 | (−0.055, 0.192) | 0.276 | ||
| Osteoporosis self-efficacy exercise | Model 1 | Race Black vs. White | 0.083 | (0.006, 0.16) | 0.036 |
| Model 2 | Race Black vs. White | 0.214 | (0.125, 0.304) | <0.001 | |
| Model 3 | Race Black vs. White | 0.213 | (0.124, 0.303) | <0.001 | |
| Intervention vs. Control | −0.028 | (−0.09, 0.034) | 0.38 | ||
| Site KP vs. UAB | 0.059 | (−0.012, 0.13) | 0.102 | ||
| Model 4 | Race Black vs. White | 0.214 | (0.124, 0.304) | <0.001 | |
| Intervention vs. Control | −0.061 | (−0.142, 0.02) | 0.142 | ||
| Site KP vs. UAB | 0.019 | (−0.077, 0.115) | 0.696 | ||
| Site (KP) * Intervention | 0.077 | (−0.048, 0.203) | 0.225 | ||
| Osteoporosis self-efficacy diet | Model 1 | Race Black vs. White | −0.01 | (−0.086, 0.067) | 0.806 |
| Model 2 | Race Black vs. White | 0.091 | (−0.003, 0.185) | 0.057 | |
| Model 3 | Race Black vs. White | 0.090 | (−0.004, 0.184) | 0.06 | |
| Intervention vs. Control | −0.031 | (−0.097, 0.035) | 0.364 | ||
| Site KP vs. UAB | 0.071 | (−0.005, 0.146) | 0.066 | ||
| Model 4 | Race Black vs. White | 0.091 | (−0.003, 0.185) | 0.057 | |
| Intervention vs. Control | −0.064 | (−0.149, 0.022) | 0.145 | ||
| Site KP vs. UAB | 0.032 | (−0.069, 0.133) | 0.538 | ||
| Site (KP) * Intervention | 0.076 | (−0.056, 0.208) | 0.258 | ||
| Patient activation | Model 1 | Race Black vs. White | −0.002 | (−0.077, 0.073) | 0.960 |
| Model 2 | Race Black vs. White | 0.117 | (0.023, 0.211) | 0.014 | |
| Model 3 | Race Black vs. White | 0.117 | (0.024, 0.211) | 0.014 | |
| Intervention vs. Control | 0.008 | (−0.058, 0.075) | 0.804 | ||
| Site KP vs. UAB | 0.002 | (−0.074, 0.079) | 0.950 | ||
| Model 4 | Race Black vs. White | 0.118 | (0.024, 0.211) | 0.014 | |
| Intervention vs. Control | −0.024 | (−0.111, 0.063) | 0.596 | ||
| Site KP vs. UAB | −0.036 | (−0.139, 0.066) | 0.487 | ||
| Site (KP) * Intervention | 0.074 | (−0.059, 0.208) | 0.274 |
Model 1 includes race only (Black vs. White)
Model 2 includes race and covariates measured at baseline
Model 3 includes race, intervention, plus covariates measured at baseline
Model 4 includes race, intervention, interaction between intervention and site (KP vs. UAB), plus covariates measured at baseline
Baseline covariates include, site (KP vs. UAB), age (continuous), education, comorbidity (COPD, depression and breast cancer), self-assessed health status (using self-reported health status, continuous), heath habits (current smoker, former smoker and drinker), bone health (history of DXA, history of fracture and history of osteoporosis treatment), health literacy (continuous), health numeracy (continuous), study DXA, fracture risk, gender of provider, physician, and provider specialty.
Discussion
The IOM defines racial disparities as “racial or ethnic differences in the quality of healthcare that are not due to access-related factors or clinical needs, preferences, and appropriateness of intervention.”12 This definition implies that before a difference can be called a disparity adjustment for patient, provider and health system factors that are difficult to measure must take place.
In an analysis of White and Black women from the PAADRN study we found that a subtle and nuanced story becomes visible after adjustment. For example, without adjustment Black women were less likely to correctly recall their DXA result, and this difference persisted after adjustment fulfilling the IOM definition of a disparity. Alternatively, without adjustment Black women appeared more likely to receive guideline-concordant therapy, but after adjustment Black women were actually less likely to receive guideline-concordant therapy, which we have labeled as a suppressed disparity. Finally, Black and White women had similar levels of patient activation in the unadjusted analyses, but after adjustment Black women actually were more activated than White women, which we have labeled as a suppressed advantage. It is important to note that these suppressed disparities and advantages were seen only after adjustment for critical clinical details—in this case, the baseline DXA results.
We are unaware of any prior studies that have evaluated the impact of educational interventions on racial differences in osteoporosis. More broadly there are relatively few RCTs assessing the impact of educational interventions on racial differences outside of osteoporosis, though a number are currently underway.22,23 Tully reported that a pilot trial of patient empowerment improved blood pressure control among Black intervention and usual care patients.24 Thomas found that an educational video reduced White-Black differences in willingness to receive an implantable cardiac defibrillator (ICD).25
A number of our other findings warrant brief elaboration. Both this analysis and related publications provide evidence that our mailed DXA result letter had similar effects in Blacks and Whites.16,19 This finding is important and suggests that despite critical differences in education, health literacy and numeracy, and baseline health status, well-crafted interventions targeting patients can provide similar benefits to both Whites and Blacks.
It is important to discuss future directions. The rise in patient advocacy, in patients being active partners in their own medical care, and in shared decision making makes it incumbent on the healthcare system to find ways to communicate results to diverse patient populations. Our tailored letters are a start, but future research needs to explore communication involving different diagnostic tests and using newer communication modalities (e.g., patient portals and text messaging with embedded YouTube links).20,21
Our study has three limitations. The first is whether these results can be applied to newer communication modalities. The second is our focus only on White and Black women at just two health systems. Finally, despite our extensive data collection we were unable to capture certain factors that are likely important in explaining racial differences, such as access to care (e.g., travel distance), provider attitudes and biases, and institutional policies.
In conclusion, using data from the PAADRN study we found that certain unadjusted differences in bone health care between White and Black women appear to be true disparities. After careful risk adjustment, we also identified suppressed disparities and suppressed advantages that were not apparent in our unadjusted analyses. These results highlight the importance of carefully planning protocols for patient (including key clinical findings), provider, and institutional data collection and subsequent statistical analyses when differentiating racial differences from racial disparities.
Acknowledgments
Funding/Support: This work was supported by R01 AG033035 (Cram/Wolinsky) from the NIA at NIH. Dr. Cram is supported by a K24 AR062133 award from NIAMS at the NIH. Dr. Saag is supported by a K24 AR052361 award from the NIAMS at the NIH.
Footnotes
Role of the Sponsor: The NIA had no role in the design and conduct of the study; collection, management, analysis, and interpretation of data; preparation, review, or approval of the manuscript; or, decision to submit the manuscript for publication.
Additional Contributions: We thank Rebecca Burmeister, MPH (UI), Mollie Giller, MPH (UI), April Miller RT (UI), CBDT, Amna Rizvi-Toner, BA, BS (UI), Brandon Van Cleave (UI), Kara Wessels, BA (UI), Roslin Nelson (KPGA), Kathy Pines (KPGA), and Brandi Robinson, MPH (KPGA), and all of the staff at the Iowa Social Science Research Center for recruiting and interviewing all study participants. All except Ms. Miller were compensated from grant funds for their time. We also thank Mollie Giller, MPH (UI), Ryan Outman, MS (UAB), Marina Reynolds, BSN (UI) Brandi Robinson, MPH (KPGA), and Jessica Williams, PhD (UAB) for coordinating and facilitating recruitment of study participants. We also thank Xin Lu, MS and Thuy Nguyen, MS (UI), for managing trial data. Finally, we thank the 7749 patients who participated in PAADRN.
Conflict of Interest: P Cram, M Jones, F Wolinsky, S Edmonds, S Hall, Y Lou, and D. Roblin have no conflicts of interest. NC Wright has received unrestricted grant support from Amgen for work unrelated to this project. KG Saag has received grants from Amgen, Eli Lilly and Merck and has served as a paid consultant to Amgen, Eli Lilly, and Merck
References
- 1.Brown CA, Starr AZ, Nunley JA. Analysis of past secular trends of hip fractures and predicted number in the future 2010–2050. J Orthop Trauma. 2012;26(2):117–122. doi: 10.1097/BOT.0b013e318219c61a. [DOI] [PubMed] [Google Scholar]
- 2.Shumway-Cook A, Ciol MA, Hoffman J, Dudgeon BJ, Yorkston K, Chan L. Falls in the Medicare population: incidence, associated factors, and impact on health care. Physical therapy. 2009;89(4):324–332. doi: 10.2522/ptj.20070107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Feldstein AC, Vollmer WM, Smith DH, et al. An outreach program improved osteoporosis management after a fracture. J Am Geriatr Soc. 2007;55(9):1464–1469. doi: 10.1111/j.1532-5415.2007.01310.x. [DOI] [PubMed] [Google Scholar]
- 4.Kaufman JD, Bolander ME, Bunta AD, Edwards BJ, Fitzpatrick LA, Simonelli C. Barriers and solutions to osteoporosis care in patients with a hip fracture. J Bone Joint Surg Am. 2003;85(9):1837–1843. doi: 10.2106/00004623-200309000-00030. [DOI] [PubMed] [Google Scholar]
- 5.Kanis JA. Diagnosis of osteoporosis and assessment of fracture risk. Lancet. 2002;359(9321):1929–1936. doi: 10.1016/S0140-6736(02)08761-5. [DOI] [PubMed] [Google Scholar]
- 6.Cauley JA. Defining ethnic and racial differences in osteoporosis and fragility fractures. Clin Orthop Relat Res. 2011;469(7):1891–1899. doi: 10.1007/s11999-011-1863-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Wright NC, Saag KG, Curtis JR, et al. Recent trends in hip fracture rates by race/ethnicity among older US adults. J Bone Miner Res. 2012;27(11):2325–2332. doi: 10.1002/jbmr.1684. [DOI] [PubMed] [Google Scholar]
- 8.Dy CJ, Lane JM, Pan TJ, Parks ML, Lyman S. Racial and Socioeconomic Disparities in Hip Fracture Care. J Bone Joint Surg Am. 2016;98(10):858–865. doi: 10.2106/JBJS.15.00676. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Screening for osteoporosis: U S. preventive services task force recommendation statement. Ann Intern Med. 2011;154(5):356–364. doi: 10.7326/0003-4819-154-5-201103010-00307. [DOI] [PubMed] [Google Scholar]
- 10.Cheng H, Gary LC, Curtis JR, et al. Estimated prevalence and patterns of presumed osteoporosis among older Americans based on Medicare data. Osteoporos Int. 2009;20(9):1507–1515. doi: 10.1007/s00198-009-0835-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hamrick I, Whetstone LM, Cummings DM. Racial disparity in treatment of osteoporosis after diagnosis. Osteoporos Int. 2006;17(11):1653–1658. doi: 10.1007/s00198-006-0178-y. [DOI] [PubMed] [Google Scholar]
- 12.Smedley BD, Stith AY, Nelson AR, editors. Institute of Medicine. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington, D.C: National Academies Press; 2003. [PubMed] [Google Scholar]
- 13.Vasquez E, Shaw BA, Gensburg L, Okorodudu D, Corsino L. Racial and ethnic differences in physical activity and bone density: National Health and Nutrition Examination Survey, 2007–2008. Preventing chronic disease. 2013;10:E216. doi: 10.5888/pcd10.130183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Wolinsky FD, Lou Y, Edmonds SW, et al. Activating Patients With a Tailored Bone Density Test Results Letter and Educational Brochure: the PAADRN Randomized Controlled Trial. J Clin Densitom. 2016 doi: 10.1016/j.jocd.2016.08.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Edmonds SW, Wolinsky FD, Christensen AJ, et al. The PAADRN Study: A design for a randomized controlled practical clinical trial to improve bone health. Contemp Clin Trials. 2012;34(1):90–100. doi: 10.1016/j.cct.2012.10.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Cram P, Wolinsky FD, Lou Y, et al. Patient-activation and guideline-concordant pharmacological treatment after bone density testing: the PAADRN randomized controlled trial. Osteoporos Int. 2016 doi: 10.1007/s00198-016-3681-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Edmonds SW, Cram P, Lu X, et al. Improving bone mineral density reporting to patients with an illustration of personal fracture risk. BMC medical informatics and decision making. 2014;14(101) doi: 10.1186/s12911-014-0101-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Edmonds SW, Solimeo SL, Lu X, Roblin DW, Saag KG, Cram P. Developing a bone mineral density test result letter to send to patients: a mixed-methods study. Patient preference and adherence. 2014;8:827–841. doi: 10.2147/PPA.S60106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Edmonds SW, Cram P, Lou Y, et al. Effects of a DXA result letter on satisfaction, quality of life, and osteoporosis knowledge: a randomized controlled trial. BMC Musculoskelet Disord. 2016;17(1):369. doi: 10.1186/s12891-016-1227-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Shultz SK, Wu R, Matelski JJ, Lu X, Cram P. Patient Preferences for Test Result Notification. J Gen Intern Med. 2015 doi: 10.1007/s11606-015-3344-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Goel MS, Brown TL, Williams A, Cooper AJ, Hasnain-Wynia R, Baker DW. Patient reported barriers to enrolling in a patient portal. J Am Med Inform Assoc. 2011;18(Suppl 1):i8–12. doi: 10.1136/amiajnl-2011-000473. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Williams KP, Roman L, Meghea CI, Penner L, Hammad A, Gardiner J. Kin KeeperSM: design and baseline characteristics of a community-based randomized controlled trial promoting cancer screening in Black, Latina, and Arab women. Contemp Clin Trials. 2013;34(2):312–319. doi: 10.1016/j.cct.2012.12.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Fiscella K, Boyd M, Brown J, et al. Activation of persons living with HIV for treatment, the great study. BMC Public Health. 2015;15:1056. doi: 10.1186/s12889-015-2382-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Tully M, Kos A, Eastwood D, Kusch J, Kotchen T. Implementation of an Adjunct Strategy to Reduce Blood Pressure in Blacks with Uncontrolled Hypertension: a Pilot Project. Ethnicity & disease. 2015;25(2):168–174. [PMC free article] [PubMed] [Google Scholar]
- 25.Thomas KL, Zimmer LO, Dai D, Al-Khatib SM, Allen LaPointe NM, Peterson ED. Educational videos to reduce racial disparities in ICD therapy via innovative designs (VIVID): a randomized clinical trial. Am Heart J. 2013;166(1):157–163. doi: 10.1016/j.ahj.2013.03.031. [DOI] [PubMed] [Google Scholar]