Racial differences in low value care among older adult Medicare patients in US health systems: retrospective cohort studyBMJ 2023; 383 doi: https://doi.org/10.1136/bmj-2023-074908 (Published 25 October 2023) Cite this as: BMJ 2023;383:e074908
- Ishani Ganguli, assistant professor1,
- Matthew B Mackwood, assistant professor23,
- Ching-Wen Wendy Yang, research programmer/analyst3,
- Maia Crawford, research project director3,
- Kathleen L Mulligan, research assistant1,
- A James O’Malley, professor34,
- Elliott S Fisher, professor235,
- Nancy E Morden, national chief medical officer, clinical program innovation and evaluation36
- 1Harvard Medical School and Brigham and Women’s Hospital, Boston, MA, USA
- 2Department of Community & Family Medicine, Geisel School of Medicine at Dartmouth, Lebanon, NH, USA
- 3The Dartmouth Institute for Health Policy and Clinical Practice, Geisel School of Medicine at Dartmouth, Lebanon, NH, USA
- 4Department of Biomedical Data Science, Geisel School of Medicine at Dartmouth, Lebanon, NH, USA
- 5Department of Medicine, Geisel School of Medicine at Dartmouth, Lebanon, NH, USA
- 6UnitedHealthcare, Minnetonka, MN, USA
- Correspondence to: I Ganguli
- Accepted 14 September 2023
Objective To characterize racial differences in receipt of low value care (services that provide little to no benefit yet have potential for harm) among older Medicare beneficiaries overall and within health systems in the United States.
Design Retrospective cohort study
Setting 100% Medicare fee-for-service administrative data (2016–18).
Participants Black and White Medicare patients aged 65 or older as of 2016 and attributed to 595 health systems in the United States.
Main outcome measures Receipt of 40 low value services among Black and White patients, with and without adjustment for patient age, sex, and previous healthcare use. Additional models included health system fixed effects to assess racial differences within health systems and separately, racial composition of the health system’s population to assess the relative contributions of individual patient race and health system racial composition to low value care receipt.
Results The cohort included 9 833 304 patients (6.8% Black; 57.9% female). Of 40 low value services examined, Black patients had higher adjusted receipt of nine services and lower receipt of 20 services than White patients. Specifically, Black patients were more likely to receive low value acute diagnostic tests, including imaging for uncomplicated headache (6.9% v 3.2%) and head computed tomography scans for dizziness (3.1% v 1.9%). White patients had higher rates of low value screening tests and treatments, including preoperative laboratory tests (10.3% v 6.5%), prostate specific antigen tests (31.0% v 25.7%), and antibiotics for upper respiratory infections (36.6% v 32.7%; all P<0.001). Secondary analyses showed that these differences persisted within given health systems and were not explained by Black and White patients receiving care from different systems.
Conclusions Black patients were more likely to receive low value acute diagnostic tests and White patients were more likely to receive low value screening tests and treatments. Differences were generally small and were largely due to differential care within health systems. These patterns suggest potential individual, interpersonal, and structural factors that researchers, policy makers, and health system leaders might investigate and address to improve care quality and equity.
Many studies have shown that Black patients in the United States and elsewhere are less likely to receive high value healthcare than White patients,123456789 but the evidence is less clear on racial differences in receipt of low value care10111213—services that provide little to no benefit in specific clinical scenarios yet have potential for harm.1011 Studies exploring a small set of low value services (eg, cervical cancer screening in older women) have found no differences or greater receipt among White patients compared with patients from racial and ethnic minority groups.101113 Given financial, physical, and other harms of low value care use,14 which has decreased marginally or not at all in recent years despite substantial attention,1516 the limited evidence on racial differences in this care represents a barrier to addressing health and healthcare inequities faced by Black people in the United States and improving health outcomes overall.1718
Racial differences in low value care receipt could be influenced by several factors. Firstly, health systems—where a growing share of Americans receive care—might shape low value care use through clinical policies, workflows, investments, hiring practices, and quality measurement.1920 As a result of such influences, researchers have noted differences in overall care quality both within and across health systems; that is, Black and White patients have been shown to receive differential care due to bias within systems and because Black patients disproportionately receive care at lower quality systems.12392122232425 Secondly, racial differences in low value care receipt might be influenced by broader differences in healthcare use; for example, differential access to primary and specialty care resulting, at least in part, from structural racism.262728 Understanding racial differences in low value care receipt overall and within systems is critical for developing interventions to reduce low value care use and promote equity.
To address this need among older adults who are at high risk of low value care,151929 we used 100% Medicare fee-for-service claims data to compare Black and White older adults attributed to 595 US health systems on receipt of 40 low value services. These services fell into four distinct categories: screening tests, acute diagnostic tests, monitoring tests, and treatments. We then built on these comparisons by investigating whether racial differences in low value care use persisted when comparing Black and White patients within the same systems, and by examining the relative contributions of individual patient race and health system racial composition to differences in low value care receipt. We hypothesized that racial differences would vary across the four categories of low value services studied, and that these patterns might generate causal hypotheses and inform future interventions.
Data and population
We used 2016-18 Medicare fee-for-service administrative data from the Master Beneficiary Summary file; Medicare Part A (hospital) and B (outpatient) 100% claims; Medicare Part D (prescription) claims for a random 40% sample of patients; the Long Term Care Minimum Data Set; and FirstDataBank.30 Our study cohort included patients aged 65 and older as of 1 January 2016 and continuously enrolled in Medicare Parts A and B through 2018 or until death. For measures using prescription information (for outcomes or for low value care algorithms or exclusions), we required continuous Part D enrollment through 2018 or until death. We excluded patients with any hospice use in 2017-18 (Master Beneficiary Summary File—MBSF) and those not attributable to a health system. We limited our analyses to non-Hispanic Black and non-Hispanic White patients, as defined by the Research Triangle Institute developed race variable (MBSF).31 We did not assess patients in other racial groups to focus on the unique experience of Black Americans and because of limitations in identifying other groups using claims data.31
Identifying health systems and attributing patients to systems
We identified health systems listed in the 2018 Agency for Healthcare Research and Quality Compendium.1932 We attributed each patient with at least one clinician encounter (ie, an evaluation and management (E&M) service, using Carrier and Outpatient files) to a single health system based on the plurality of primary care services received across 2017 and 2018 following Centers for Medicare and Medicaid Services (CMS) Medicare Shared Savings Program attribution methods.33 We excluded systems that were predominantly pediatric (based on a compendium indicator for system hospitals primarily serving children). We further restricted our analysis to systems with ≥250 attributed patients.
Low value care measures
We operationalized 40 claims based measures of low value care relevant to older adults from previous research19 and from the Milliman MedInsight Health Waste Calculator version 8.0,34 a standalone, proprietary, annually updated software program that identifies potentially low value services based on Choosing Wisely campaign recommendations and other professional medical society guidelines (supplementary table S1).1519 Results from the Milliman MedInsight Health Waste Calculator have been used by a number of insurers, employers, and state governments to inform policies and have been published in peer reviewed journals.151935363738 Milliman algorithms use international classification of diseases, ninth and tenth revision (ICD-9 and ICD-10) diagnosis codes; American Medical Association procedural codes; and National Drug Code entries to assign healthcare services provided within specific clinical scenarios to one of three categories: not wasteful, likely wasteful, or wasteful. In our analyses, we defined low value services conservatively by only including those tagged as wasteful. For each measure, we used 2016-18 data to identify patients eligible (in other words, at risk) for the low value service based on claims derived demographic and clinical characteristics including age, sex, previous medical conditions, and previous service use. For instance, female patients older than 65 with adequate previous screening and not otherwise at high risk for cervical cancer would be considered eligible for low value cervical cancer screening. For each service, we then calculated the share of eligible patients who received the low value service at least once in 2017-18. For ease of interpretation, we assigned each service to one of four categories based on service type, indication, and clinical context39: screening tests, acute diagnostic tests, monitoring tests, and treatments.
Patient and system factors
To account for individual level factors that might confound racial differences in low value care use, we obtained patient age (continuous, 2016 MBSF), sex (2016 MBSF), and number of ambulatory visits in 2016 (E&M codes 99201-5, 99211-5 in Carrier and Outpatient files). We also identified patients’ Medicaid eligibility status, CMS Hierarchical Condition Category (HCC) score (a measure of medical complexity, continuous, 2016 Medicare administrative data40), share of patients’ outpatient visits billed by a primary care clinician (doctors specializing in general practice, family practice, internal medicine, pediatric medicine, or geriatric medicine according to CMS methods; we also included nurse practitioners, certified clinical nurse specialists, and physician assistants because some have primary care roles—though Medicare data do not capture specialties for these clinicians), and continuity of care (2016, Bice-Boxerman Continuity of Care Index41; range=0 (each visit with different clinician) to 1 (all visits with same clinician)). Finally, we examined the racial composition of health system populations (defined as the percentage of all patients attributed to the health system who were Black) and safety net status (generally meaning patients are cared for regardless of ability to pay; here defined by health system inclusion of at least one hospital with a high Disproportionate Share Hospital patient percentage42; AHRQ Compendium; supplementary table S2).
We estimated unadjusted rates of each low value service among Black and White patients. To estimate differences by race in the receipt of each service, we then constructed our primary models: a series of patient level, multivariable, logistic regression models in which the key predictor was patient race and additional covariates were patient age, sex, and ambulatory care use, with standard errors clustered at the health system level using Huber-White correction. In these primary models, we did not adjust for factors such as HCC score (low value service measures already account for clinical factors) or Medicaid eligibility (an indicator of poverty) according to National Academy of Medicine guidelines to avoid masking racial differences.4344 Then, in secondary analyses, we repeated these models for each low value service, additionally including health system fixed effects to compare Black and White patients within the same systems. Comparing the results of models without and with system fixed effects allowed us to assess the extent to which racial differences in service receipt persisted within the given systems. If differences between Black and White patients were similar in both sets of models, this would suggest that differences were largely explained by differential care within systems; if differences diminished when adding system fixed effects, these differences might be better explained by Black and White patients receiving care from different health systems.
We performed two sets of exploratory analyses that were modifications of our primary models. Firstly, to determine if racial differences in low value care receipt diminish when comparing Black and White patients with similar socioeconomic status, we added Medicaid eligibility to these models. Second, to explore intersectional effects of race and sex,45 we added race-sex interaction terms to the models for the 36 services that male and female patients could receive (eg, we excluded prostate specific antigen testing).
Then, to determine the relative roles in low value care receipt of the patient’s race versus the racial composition of the patient’s health system, we built three more sets of models. The first set included patient race, age, sex, ambulatory care use, and health system random effects (to account for unobserved differences between health systems while also allowing us to estimate the effects of the health system level factors we describe next). The second set also included an indicator for racial composition of the patient’s attributed health system. The third set of models included indicators for racial composition and safety net status of the health system because systems with this status might have greater experience with caring for underserved racial and ethnic minority groups.
Finally, to assess the extent to which differences by individual race varied across health systems after accounting for the racial composition of systems, we built a set of models that included individual race, system racial composition, the above covariates, health system random intercept effects, and health system random slopes for individual Black race. The variance of these random slopes quantifies the unexplained variation between systems in individual Black-White differences.
We used complete case analysis because there were no missing values for model covariates. We used postestimation counterfactual evaluations to translate logistic regression results to percentage point differences for ease of interpretation. We created claims measures and ran descriptive and regression analyses using SAS, version 9.4 (SAS Institute) and Stata, version 17.0 (StataCorp LLC). To create the figures, we used R in RStudio. Reported P values were two sided and P<0.05 represented statistical significance. For the primary models, we used Holm-Bonferroni correction to adjust for multiple comparisons.46 This study followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting.
Patient and public involvement
No members of the public were formally involved in the design or implementation of this study. However, clinician authors’ lived experiences as members of racial and ethnic minority groups and their clinical experiences interacting with minoritized patients informed the research question and interpretation of results.
We analyzed data for 671 890 Black patients and 9 161 414 White patients attributed to 595 health systems (table 1). Black patients were more likely to be female (62.6% v 57.6%) and Medicaid eligible (17.4% v 4.8%) compared with White patients. On average, Black patients were younger, had higher HCC scores, had a higher proportion of ambulatory visits with primary care physicians or advanced practice clinicians (55.6% v 50.0%), and had lower Bice-Boxerman Continuity of Care scores (0.17 v 0.20). Of the 595 systems, 204 (34.3%) had safety net status. Systems varied widely in the proportion of attributed patients who were Black, ranging from 0.07% to 82.0% (median 3.7%).
Racial differences in low value service use by service type
Unadjusted rates of low value service use varied by race across measures (fig 1; supplementary table S3). Supplementary table S4 shows numbers of Black and White patients receiving each service. Adjusted results were similar to unadjusted results. In adjusted analyses, Black and White patients differed significantly in receipt of 29 low value services (fig 2; supplementary table S3). Black patients had higher rates of nine low value services: one screening test, six acute diagnostic tests, and two treatments. White patients had higher rates of 20 low value services: nine screening tests, one acute diagnostic test, two monitoring tests, and eight treatments. Specifically, Black patients were at greater risk than White patients of receiving acute low value diagnostic tests, including imaging for uncomplicated headache (6.9% v 3.2%), emergency department head computed tomography scans for dizziness (3.1% v 1.9%), and head imaging for syncope (2.6% v 2.1%). Black patients were also more likely to be prescribed two or more antipsychotic drugs (8.0% v 5.2%) and to receive a feeding tube in the setting of advanced dementia (8.5% v 2.2%).
In contrast, White patients were at higher risk of receiving low value screening services, including preoperative laboratory testing (10.3% v 6.5%); electrocardiograms and other cardiac screens (5.1% v 1.8%); and prostate specific antigen testing in men over age 75 (31.0% v 25.7%). White patients were also more likely to receive low value treatments, including antibiotics for acute upper respiratory and ear infections (36.6% v 32.7%), renal artery revascularization (4.5% v 2.3%), injections for low back pain (5.6% v 4.2%), and vertebroplasty for osteoporotic fractures (4.9% v 2.8%).
When we added Medicaid eligibility (a proxy for poverty) to the primary models, the results did not change appreciably for most services with a few exceptions: differences diminished for some services that White patients were more likely to receive than Black patients—for example, antibiotics for upper respiratory infection and prostate specific antigen testing (supplementary table S5). When examining race and sex interactions, 14 low value services showed statistically significant interactions (supplementary table S6). For example, Black female patients were more likely to receive imaging for uncomplicated headache (7.5% v 3.4% for White female patients, 5.5% for Black male patients, and 2.8% for White male patients); and White female patients were more likely to have electrocardiograms and other cardiac screens (6.1% v 3.9% for White male patients, 1.9% for Black female patients, and 1.8% for Black male patients) and antibiotics for upper respiratory infections (38.6% v 33.0% for White male patients, 36.0% for Black female patients, and 26.0% for Black male patients).
Racial differences in low value service use across versus within health systems
When we added system fixed effects to the primary models, the direction and magnitude of the differences were similar, meaning that racial differences in low value care receipt were largely due to differential care within systems rather than Black and White patients receiving care in different systems (fig 2). We noted a few key exceptions. Firstly, Black patients were more likely to receive imaging for eye disease than White patients, and this difference widened when comparing Black and White patients within the same systems. Secondly, Black patients were less likely to receive low value prostate specific antigen testing and cervical cancer screening, and these differences also widened when comparing Black and White patients within the same systems.
In models including system random intercepts and slopes, the direction and magnitude of individual racial differences were largely unchanged when also accounting for racial composition of the patient’s attributed system (table 2), consistent with system fixed effect model results reported above. However, even after accounting for individual race, patients attributed to systems serving a larger share of Black patients received more of certain services, including low value repeat bone density testing (0.74 percentage points for every 10% of Black patients served, 95% confidence interval 0.40 to 1.08) and prostate specific antigen testing (1.18 percentage points, 0.46 to 1.90), and less of other low value services, including radiation for bone metastases (−0.28 percentage points, −0.46 to −0.10). When we also included system safety net status in the models, our results were largely unchanged (supplementary table S7). Racial differences in low value care receipt varied widely across health systems (supplementary table S8).
In this analysis of Medicare patients attributed to US health systems, a range of <1% to 37% of eligible patients received each of the 40 low value services examined. For most of these services, we found significant racial differences, ranging from modest (most differences) to more than fourfold, with notable patterns by service type. Black patients were more likely to receive some low value acute diagnostic tests such as imaging for uncomplicated headache, syncope, and low back pain, while White patients were more likely to receive most low value screening tests and treatments, including preoperative tests, cardiac and cancer screening, and surgical procedures. The observed racial differences were largely driven by differential care within systems rather than by Black and White patients receiving care from different systems.
Strengths and weaknesses
This study examines racial differences in low value care use across a large number of low value services with a range of impacts on spending and patient harms,39 and examines the role of health systems in these differences. These results build on previous research that found reduced risk of low value cervical cancer screening among Black US veterans13 and Medicare patients.11 Similar to the findings of Schpero and colleagues (2006-11 data),11 we found Black patients with advanced dementia were at higher risk of receiving feeding tubes. However, unlike Schpero, we found that Black patients experienced no significant difference in vitamin D screening, lower risk of antipsychotic use in dementia, and higher risk of imaging for low back pain.
The study has several limitations. The 40 services examined represent a fraction of all low value care, and claims data lack clinical details to confirm clinician intent.47 Potential for bias exists if misclassification of low value care eligibility varies systematically by race. We also acknowledge important methodological debate in choice of model covariates.4344 Our data are from 2016 to 2018, and differences in low value care use might have changed since then. The study focuses on patients attributed to larger US health systems, therefore our results might not generalize to patients attributed to small systems or to those who are not attributed to health systems—who might receive low value care at slightly higher rates.48 However, we note that Black and White patients had similar likelihood of being attributed to the included systems (when applying our cohort criteria to Medicare patients before and after system attribution, Black patients made up 6.9% of all patients and 6.4% of all attributed patients). Similarly, our results might not generalize to the growing share of older adults enrolled in Medicare Advantage or to other populations.29
Our intention in this study was to generate hypotheses; underlying mechanisms should be explored in future research. We did not involve the public in shaping study questions and interpretation, which will be critical for future studies. Future quantitative and qualitative work might also assess low value care receipt among other racial groups, examine the role of health system characteristics in inequities, explore the association between high value and low value care, and study positive outliers in net care value to identify best practices.
Although the racial differences we found were heterogeneous and generally modest, they varied by low value service category in informative ways. These patterns suggest hypotheses about possible mechanisms at individual, interpersonal, and structural levels that can inform low value care reduction efforts more broadly.
Black patients were at slightly (one to twofold) greater risk of receiving several low value acute diagnostic tests. At the individual level, mistrust in the healthcare system because of historical and present day racism4950 might contribute to Black adults being more receptive to diagnostic testing when acutely ill—in this scenario, it is possible that a tangible test is more reassuring than a clinician’s words and might serve to lessen valid concerns about undertreatment. Mistrust, and structural factors such as access barriers, might also result in care seeking delays, so that Black adults are sicker when they present with acute conditions, potentially leading clinicians to order more low value tests.
At an interpersonal level, clinicians’ implicit or explicit biases51 and patient-clinician racial discordance5253 might contribute to clinician misperceptions of patient needs, less effective communication, and in turn, greater clinical uncertainty54 that prompts more testing.555657 In a similar vein, we found Black patients with advanced dementia were more likely to receive feeding tubes, mirroring evidence that patients from racial and ethnic minority groups were more likely to receive high cost, aggressive interventions at the end of life.255859 These findings may be due to patients in racial and ethnic minority groups having fewer treatment limitations in place (eg, do not resuscitate orders), more severe illness, or different end-of-life preferences.5758 At a structural level, these results might reflect decades of structural racism resulting in racially segregated neighborhoods and lower density of primary care clinics and high quality urgent care centers60 in predominantly Black neighborhoods compared with White neighborhoods. Despite our finding that Black patients had a larger share of visits with primary care or advanced practice clinicians, they are less likely to have a regular primary care clinician2861 (who might appropriately triage them away from the emergency department) and might be more likely to receive acute care from urgent care or emergency medicine clinicians who do not know them well,2762 potentially contributing to more acute diagnostic testing.63 This low value acute diagnostic testing affects recipients through direct harms (eg, radiation exposure, treatment burden, and out-of-pocket spending)64 and through downstream care cascades that can often arise from imaging tests in particular.6566
In contrast to our finding that Black patients received more low value acute services, White patients received one to threefold more of most low value screenings examined. At the individual level, it is possible that White patients are more likely to request or agree to screening services if relative privilege and the racial opportunity gap augment their perceived benefit of such services compared with Black patients.67 At an interpersonal level, clinicians might be more likely to offer screening services to White patients because of implicit biases. Some racial differences could be explained by socioeconomic differences, as suggested by our finding that the greater risk of low value prostate specific antigen testing in White men compared with Black men was diminished when controlling for a poverty indicator. At a structural level, Black patients have less access to routine, timely primary care (despite our finding that Black patients had a larger share of visits with primary care doctors or advanced practice clinicians).22286168 For instance, Black Medicare patients are less likely to receive annual wellness visits2268 and use disproportionately more emergency department and inpatient care relative to ambulatory care.2769 Unfortunately, lower rates of low value screening among Black patients parallels lower rates of high value screening, such as age appropriate colorectal cancer screening.70 Most notably, higher rates of prostate cancer incidence and death among Black men71 could arguably justify higher rates of low value prostate cancer screening in this population, but Black men were instead less likely than White men to receive this service. Finally, White patients were at slightly (one to threefold) higher risk of receiving eight of 12 examined low value treatments, such as antibiotics for acute upper respiratory tract and ear infections. These findings might reflect White patients being more likely to request treatments, or clinicians being more inclined to offer them to White patients, partly because of implicit, historically rooted power differentials.7273
Our exploratory examination of race and sex interactions revealed some substantial differences in low value care receipt across sex-race groups. Hypotheses that might explain these patterns would be speculative, but we believe these findings merit further examination in future studies.
Examining the role of health systems, we found racial differences in receipt of low value care were driven less by Black and White patients receiving care in different systems, and more by differential treatment within systems (eg, through services directly provided by in-system primary care and specialist clinicians, or through referrals19), consistent with previous evidence of stronger within-system racial inequalities in care quality measures.9 The most notable exception to this was prostate specific antigen testing—health systems serving larger proportions of Black patients offered more low value prostate specific antigen tests to all of their patients, perhaps in response to higher rates of prostate cancer among Black men. However, within any given system, White patients were still more likely to get the service. Studies have revealed that hospitals or practices that predominantly serve patients from racial and ethnic minority groups provide lower care quality for all of their patients,212223 but we did not find clear evidence of this in our study.
Low value care is problematic, not only because of direct and cascading harms to patients,1439 but also because it diverts limited resources, contributing to underuse of effective care especially among racial and ethnic minority populations. In general, we found Black patients were at modestly greater risk of receiving low value acute diagnostic tests commonly performed in acute care settings, while White patients were at modestly greater risk of receiving low value screening services and treatments. Even small differences might be clinically important because of the direct and indirect effects of low value care, while the patterns across service categories suggest hypotheses about underlying mechanisms. We also found that these differences result almost entirely from differential care delivered within systems. Taken together, these results invite further exploration of differential access by race to routine, high value primary care, patient-clinician concordance, and trust. Our findings also highlight the need to develop and test effective interventions to reduce low value services, especially those with greatest potential impact (ie, based on numbers of people affected, direct and cascading costs, and likelihood of other harms).39 Health systems invested in reducing low value care and promoting health equity could begin by measuring low value service use internally overall, and by race and sex, in line with a recent Joint Commission mandate to report care quality data stratified by demographic categories.74 Such efforts might allow systems to identify and address underlying sources of racial differences; for example, clinician-patient interactions (bias, mistrust) or structural issues (access to high quality primary care or differential referral patterns). Granular, population stratified analyses within health systems are increasingly feasible with the use of electronic health record data, and will likely be essential to advancing equitable, high value care for all.
What is already known on this topic
Black patients in the United States are less likely than White patients to receive high value health care
Evidence on racial differences in low value care—services which provide little to no benefit yet have potential for harm—is scant and mixed
Studies showed that among a limited number of low value services, Black patients had lower rates of some services than White patients and similar rates of other services
What this study adds
Older Black Americans were more likely to receive low value acute diagnostic tests than older White Americans; older White Americans were more likely to receive low value screening tests and treatments
These differences were generally modest and were largely driven by differential treatment within health systems rather than by Black and White patients receiving care from different health systems
The results highlight the need for health systems to track internal data by race on low value care to identify, understand, and address the sources of racial differences
The study was approved by the Committee for the Protection of Human Subjects at Dartmouth College (study No 31812).
Data availability statement
Data presented in this study were obtained from US governmental agencies and vendors. We are not allowed to share these data due to existing data protection rules and licensing agreements.
We thank Benjamin Usadi, formerly of The Dartmouth Institute, for his contributions to this manuscript.
Contributors: IG, ESF, MBM, and NEM contributed to the design of the study. IG, MBM, AJO, ESF, and NEM directed the analyses, which were carried out by C-WWY. All authors contributed to interpreting the results, drafting the manuscript, and critically revising the manuscript for important intellectual content. IG, MC, and ESF obtained funding for the study. All authors approved the final version for submission. IG is the guarantor. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.
Funding: This work was supported by grant 19-02548 from Arnold Ventures and grant K23AG068240 from the National Institute on Aging. This work was also supported in part by the Agency for Healthcare Research and Quality’s Comparative Health System Performance Initiative under grant 1U19HS024075, which studies how health care delivery systems promote evidence based practices and patient centered outcomes research in delivering care. The statements, findings, conclusions, views, and opinions contained and expressed in this article are based in part on data obtained under license from IQVIA information services: OneKey subscription information services 2010-20, IQVIA Inc all rights reserved. The statements, findings, conclusions, views, and opinions contained and expressed herein are not necessarily those of IQVIA Inc or any of its affiliated or subsidiary entities.
Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/disclosure-of-interest/ and declare: support from Arnold Ventures, the National Institute on Aging, and the Agency for Healthcare Research and Quality for the submitted work; IG reports receiving consultant fees from F-Prime Capital; NEM is employed by United HealthCare, which played no role in the development or publication of this paper; no other financial relationships with any organizations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influence the submitted work.
The lead author (the manuscript’s guarantor) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.
Dissemination to participants and related patient communities: The findings of this study will be shared publicly through lay press coverage, social media, press releases from the authors’ affiliated organizations, and presentations at virtual and in-person conferences.
Provenance and peer review: Not commissioned; externally peer reviewed.
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.