Provider Participation in a National VA Pragmatic Trial: The Diuretic Comparison Project

Introduction

There is growing recognition of the utility of stakeholder engagement in clinical trials. Concannon et al. identified healthcare providers as one of the key stakeholders in comparative effectiveness research, and recognized the need for research to understand their effect on study processes such as patient recruitment.¹ More specifically, patients who receive clinical trial information from their own primary care providers are more likely to participate, compared to those who receive it from newly introduced research staff.² This may be attributed to the established trust patients have with their primary care providers. While providers can enhance patient participation in clinical research through information exchanged at the point of care, provider attitudes towards the study are also important. Fogel and colleagues found that provider views of a clinical trial can influence patient trust towards a study.³ Providers who believe that a study could be of benefit to their patients or to the healthcare community may be more motivated to offer information and encourage patient participation.

Unfortunately, harnessing the support of providers is not always straightforward. Ngune and colleagues stressed the importance of enlisting providers in research without assigning them too much responsibility beyond usual care.⁴ For example, in a review of 70 trials, only 12.5% of studies achieved recruitment targets when general practitioners were charged with obtaining patient consent. By contrast, 61.5% of studies met the targets when consent was obtained by research staff.⁵ Barriers to physician support include burdens like patient education, extensive paperwork, and additional clinic visits.⁶ These factors highlight the value of clinical research studies with pragmatic designs, which minimize additional burden on providers through integration of study procedures with usual care.

The Diuretic Comparison Project (DCP) was a pragmatic randomized trial conducted within the Veterans Affairs (VA) Healthcare System that compared the effects of two thiazide-type diuretics, chlorthalidone and hydrochlorothiazide, on cardiovascular outcomes. Study information is summarized in Figure 1, and the full DCP study design and complete findings are published elsewhere.^7,8 Primary care providers (PCPs) were not considered to be conducting research in DCP, but were instead considered participants. Invitations to join the study were sent through the computerized entry system as electronic health record (EHR) orders. Accepting the order served as a PCP’s consent to participate in the study, thus allowing the study team to pre-screen, contact, and consent patients in their panels. Consented providers were responsible for approving or declining EHR orders to randomize their individual consented patients. Randomization involved a 1:1 chance of assignment to either continue hydrochlorothiazide or switch to chlorthalidone. Once patients enrolled, their primary care providers proceeded with usual care, including management of the patient’s hypertension and study medication.

The DCP design relied on provider participation to initiate patient consent and randomization workflows, permitting study staff to prescreen patients from the assigned panels and randomize eligible consented patients. Understanding provider participation in the DCP holds value for future pragmatic trials, including those in development at the VA. Using DCP study data, this paper characterizes provider participation by reporting relevant study metrics and summarizing themes expressed by consented providers about their experiences and decisions concerning the study. In addition, we evaluate factors that both facilitated and inhibited provider participation.

Methods

This mixed methods study performed an in-depth review of provider participation in the DCP using both provider and patient-level data, semi-structured interviews with providers at one study site, and a survey disseminated to providers at all study sites. These activities were conducted under the study’s VA IRB approval.

DCP study data

DCP study data regarding provider and patient enrollment were summarized with frequencies and relative frequencies. Our metrics of interest using provider data were the overall provider consent rate and the rates at individual sites, as well as the percentage of providers who declined randomization of consented patients. Metrics of interest for patient-level data included the overall and site-specific randomization rate, along with the percentage of consented patients declined from randomization by their provider.

Provider interviews

Following the close of patient recruitment in fall of 2021, 103 consented PCPs at the Minneapolis VA Healthcare System were invited to participate in interviews about their experiences with the study. Although a total of 158 PCPs from this site consented to DCP over the course of the study, only 103 remined employed there at the time interview invitations were sent. This site was chosen because both the principal investigator and study consent call center were located there, so the local IRB was expected to be more agreeable to interviews. In addition, the site had one of the strongest provider consent rates, indicating interview recruitment might also be successful there. Unfortunately, interviews could not be conducted at additional sites due to constraints on resources and the logistical challenges of obtaining separate IRB approvals at multiple VA sites.

A semi-structured interview format was employed because it involves a pre-set list of open-ended questions, allowing interviewers to focus conversations on the topic of interest without confining participants’ responses.^9,10 A list of Minneapolis VA IRB-approved open-ended questions were asked along with follow-up questions based on participant answers. Interview questions were written to understand providers’ overall experiences and decisions surrounding their own consent and patient randomization (Supplemental item 1). Virtual interviews lasting 10-20 minutes were conducted between November 2021 and February 2022. Since the IRB approval did not cover recording, two members of the study team were present for most (75%) interviews. One led the interview while the other primarily took detailed notes. On two occasions (25%), only one interviewer was present and performed both roles. Interviews were informal in nature, and providers were given the opportunity to ask questions of their own. To enhance the effectiveness of the interviews, the interviewers employed standard techniques recommended in the literature.⁹ Following Braun and Clarke’s steps, inductive thematic analysis was performed on notes from interviews.^11,12 For each question in the interview guide, providers’ answers were compiled and coded. From these codes, themes were identified, reviewed, and refined using an iterative process until saturation was achieved.

Provider survey

In August of 2022, after DCP was completed, but prior to publication of the primary results, all consented PCPs nationwide were emailed a link to a Google Forms survey about their experiences with the trial. The survey included yes/no, multiple choice, open-ended and Likert scale questions. The full survey and results are detailed elsewhere.¹³ Questions covered a range of aspects related to the DCP provider experience and providers had four weeks to respond. Responses to selected questions, such as about the decision to randomize and study burden, are summarized herein.

Results

DCP study data

Figure 2 summarizes statistics on DCP provider and patient enrollment. Over the course of the DCP, 4,128 (69%) of 6,010 total invited providers consented. Across individual study sites, provider consent rates varied widely (ranging from 46% to 90%). Among consented providers, 14% declined at least one patient randomization. Of this cohort, 58% declined only one randomization order, while 42% declined two or more patients from randomization. One provider was an outlier, declining 21 patient randomizations. The remaining providers declined 13 or fewer randomizations. Over 13,500 (82%) of the 16,595 consented patients were successfully randomized. Some (12%) of the consented patients could not be randomized due to ineligibility or withdrawal of consent. However, 7% of all consented patients were not randomized due to their PCP declining randomization. Randomization rates also varied widely by study site (ranging from 47% to 91%), with an average site randomization rate of 80%. At the Minneapolis VA, where interviews were conducted, 85% (n=158) of invited providers consented to the study (accounting for 3.8% of all consented DCP providers), and 88% (n=737) of consented patients were randomized (5.4% of randomized DCP patients).

Provider interviews

Eight consented providers at the Minneapolis VA responded and were interviewed. Five of them had received randomization orders. Among these five, two recalled declining a randomization, and each had declined two randomizations.

Interview themes are summarized in Table 1. When asked about their decision to participate in the DCP, valuing clinical research was a common theme in provider answers. The study’s relevance to primary care was another top factor. Notably, three providers explicitly named the late Dr. Frank Lederle (the original principal investigator) as a motivating factor. Popular themes that emerged with regard to approving randomizations included medication interchangeability and a desire to support the study. When it came to declining randomizations, multiple responses fell under the themes of preserving patient well-being and the patient not being an ideal research candidate. Two interviewed providers could not name a reason they would decline to randomize a patient. On the topic of overall study experience, responses were generally positive, with nearly every provider mentioning that the study was low burden on them. Finally, prolonged study duration and providers expressing difficulty remembering details about their experience were two recurring and interrelated concepts.

Provider survey

The provider survey yielded 180 responses, representing about 4% of consented providers. 6% (10) of the respondents acknowledged declining one or more randomizations. Eight providers reported declining one patient from randomization while two providers reported declining two or more. Selected survey results are summarized in Table 2.

Those who declined randomizations were asked to share reasons why they did so. Four providers selected “I was concerned that changing my patient’s diuretic would cause harm to their health,” and two separate providers wrote-in that they were specifically worried about chlorthalidone causing or exacerbating hypokalemia. In addition, three separate providers selected “I did not wish to change my patient’s diuretic,” suggesting they were satisfied with how their patient was doing on hydrochlorothiazide and saw no reason to make unnecessary changes. One provider wrote that they declined randomization so they could switch the patient over to chlorthalidone. In one case, a provider wrote-in that they declined on the basis that the patient is not compliant, and therefore presumably not a good study candidate.

Discussion

Nearly 70% of providers who were asked to participate in DCP agreed to do so. While this is a strong showing compared to PCP participation rates reported by some other studies,^14,15 there is still room for improvement. Nearly a quarter of providers invited to participate in DCP declined. In addition, 8% of providers who were sent requests did not respond, a disappointing percentage because the process to reply was electronic and therefore low pressure and fast. Each provider received the invite up to three times, ensuring ample opportunity to respond but minimizing healthcare interruptions. Since patients could only be recruited from the panels of consented providers, each provider who declined or did not answer limited the pool of potential participants. Early in the study, the team learned that some providers were confused by the study invitations because they looked like the screens used in EHR trainings and system tests. An explanatory addendum was attached to the invitation; however, it remains possible that confusion contributed to these statistics.

Our interviews revealed a possible strategy for generating provider enthusiasm for future studies. A few interviewees attributed their involvement to Dr. Frank Lederle, a well-liked provider and esteemed cardiovascular researcher who was passionate about pragmatic research. Early on, PCPs were drawn to the study by his involvement. After his death in 2018, providers were inspired to join DCP in his memory. Dr. Lederle’s influence was concentrated within the Minneapolis VA study site, which had some of the highest provider consent and randomization rates and contributed the highest raw number of providers and randomized patients of any site. While the extent of the “Lederle effect” on DCP provider participation is not quantifiable, his legacy highlights how trusted physician-leaders can shape their local clinical research culture. Indeed, Johnston et al. cite the importance of local “opinion leaders” who have sway with colleagues and will “champion” a study.¹⁴

Differences in rates of provider participation among sites can also be a consequence of how facility leadership influences the research culture at a site. Research participation can be used as a performance metric for individual VA healthcare facilities. At sites like these, the DCP’s low-burden design may have appealed to providers looking to increase their research involvement. On the other hand, some sites were less receptive to the DCP during site recruitment, and this reluctance from leadership may have impacted provider consents.

Over 80% of patients who consented to DCP were randomized. Unfortunately, 7% of patients who consented (over one-third of non-randomized patients) were declined from randomization by their providers. Declined randomizations were frustrating for the study team, as time and resources were invested into consenting the provider, then the patient, confirming eligibility, and then contacting the provider for approval. Since there was no mechanism for PCPs to explain why they were declining a randomization, multiple declines might indicate that a provider was having issues with the study, or that a consented provider was on leave and a covering provider was declining randomizations. However, of 14% of consented providers who declined a randomization, the majority of these declined only one patient. This suggests that many providers’ decisions were patient-specific rather than reflective of issues with the study.

Survey results and interview findings support this. Provider-reported reasons for declining randomizations all had one thing in common: the decision was centered around the patient. In particular, a number expressed concern about the potential health impacts of changing their patient’s diuretic. Eligible patients were 65 and older, already taking hydrochlorothiazide for hypertension, and could have comorbidities such as diabetes, kidney disease, or other serious conditions. Providers believed that switching diuretics, despite their similarities, could have effects that might not be well tolerated by those dealing with health issues and taking other medications. Even providers who enacted a blanket policy of declining all randomization requests were motivated by patient well-being, as they expressed a belief that chlorthalidone is the superior drug for most patients and wished to switch the patient over to it rather than leaving it up to the study’s randomization. Similarly, Guillemin et al. reported that when providers had the opportunity to invite their eligible patients to participate in a clinical trial, they considered the individual patient’s suitability for research, including their risk of experiencing harm because of the study, before approaching them.¹⁶ The authors of that paper argued that this reflected providers’ conflicting obligations to the patient as their primary care provider and to the study as a clinician-researcher.¹⁶ This finding leads to a reframing of our understanding of provider participation. Providers who decline randomization of consented patients they feel do not belong in the study are in fact very engaged with the study, because they are actively considering whether each consented patient should participate.

This prioritization of the individual patient’s well-being, though positive, can also lead to physician concerns about losing control over patient care. Physicians have consistently reported discomfort that allowing patients to enroll in comparative effectiveness research will inhibit their ability to treat patients freely.¹⁷ While this misconception was not raised in our interviews or surveys, it was a concern expressed to the study team during the study. Future studies using the pragmatic design should emphasize that providers will continue to manage their patients’ care—including medications—after they are enrolled in the study. In the case of DCP, this meant that providers could change the patient’s diuretic if they felt the assigned drug was not the right fit.

The team had considered that declined randomizations might be the result of provider characteristics, such as provider inexperience; study-based reasons, like additional perceived workload; and external factors, like the COVID-19 pandemic. These possibilities are not supported by survey and interview results. Seventy-eight percent of survey respondents, and 90% who declined a randomization, had been practicing for ten or more years.¹³This data does not disprove the role of experience. However, previously, both age and years of practice have not been found to relate to interest with participating in clinical trials.¹⁸ Over 80% of all respondents and respondents who declined a patient randomization in DCP rated the study activities as either “not at all burdensome” or “not very burdensome,” and multiple providers remarked on the ease of the study during interviews. To prevent study-specific alert fatigue, study nurses had limited the number of alerts and orders to four per week per provider. In addition, most orders only required a signature, minimizing the number of “clicks.” Eighty percent of respondents who declined a randomization reported that they were either equally or more comfortable with participating in DCP during the COVID-19 pandemic,¹³ and interviewed providers echoed this sentiment, saying their attitude towards the study did not change during COVID-19.

The DCP had strong provider participation throughout the study lifecycle. Part of the study’s success with provider participation can be attributed to study design and implementation. As a usual care study, there were no mandated extra clinic visits, tests, or added paperwork. The whole process for consenting providers and enrolling patients was designed to minimize burden on PCPs. In particular, pacing study alerts was likely key to maintaining provider participation beyond the consent stage.

Limitations

Low levels of provider responsiveness to both the survey and interview opportunities were a limitation for this paper. Only about 4% of consented providers completed the provider experience survey. Only 8 of 103 Minneapolis providers agreed to be interviewed. Furthermore, interviews were only conducted at one site. The colocation of interviewed PCPs and the study’s principal investigator likely impacted the responses given.

The marketing of the DCP as a low burden study, which was intended to incentivize participation in the study itself, may have disincentivized participation in the survey or interviews, because providers felt they had fulfilled their responsibilities to the study. Additionally, because of the study’s extended duration, many providers likely had forgotten about it by the time they were sent the survey link and interview invitation. Comments from interviews support this.

While the survey and interviews help to contextualize the DCP provider participation statistics, it is important to recognize that both methods elicited self-reported data from providers. Therefore, it is subject to the potential response bias inherent in this kind of data. It is possible that non-patient-centered factors had an impact on provider’s decision-making but could not be captured using these methods.

Finally, the DCP started recruiting providers and patients in 2015 and continued enrollment until the end of 2021. After the initial wave of consents from a provider’s panel tapered, providers might go months to years without hearing from the DCP team until another patient from their panel was approached and consented. Both interviewed and surveyed providers acknowledged having forgotten about the DCP or their own involvement in it due to the long duration. It is possible that some providers received randomization orders and declined because they had simply forgotten what the DCP was and what approving the order would mean for themselves and their patients. This highlights the importance of maintaining regular communication with consented providers throughout the life of the study. Mahmud et al. found that providers felt regular meetings to review trial protocols would improve provider engagement with trials.⁶

Conclusion

Providers had high levels of participation with the DCP at the crucial stages of consent and randomization. By contrast, providers were not very responsive to survey and interview requests following the study. Providers consented to the DCP because they value clinical research, saw the applicability of the study question to their work, and wanted to support a local colleague. Declined randomization orders resulted from providers’ knowledge and concern for well-being of the consented patient and did not fully reflect a provider’s level of participation with the DCP as initially thought. Logistical and organizational challenges that typically impede provider support of clinical trials were avoided by DCP’s pragmatic design. Future pragmatic studies will be able to build upon the lessons learned from the DCP:

• identify and harness site-specific advantages such as physician-leaders or policies that promote research
• tailor provider education materials to emphasize that the study will not inhibit their ability to manage patients’ care and make best decisions for the patient
• pace study-related orders to prevent provider burnout
• regular communication with consented providers to keep the study fresh in their minds

Olivia Taylor*,¹ BA; Eustacia Ikeri,¹ BS, MPH; Alison Klint,² MS; Sarah Leatherman,^2,3 PhD; Amanda Guski,² MA; Liang Zhu,² PhD; Cynthia Hau,² MPH; Patricia Woods,² RN, MSN; Colleen Hynes,² BSN; Christal Sadatis,² BSN; Maura Flynn,² NP; Peter Glassman,⁴ MBBS, MSc; Addison Taylor,⁵ MD; Srihari Raju,¹ MD; William Cushman,⁶ MD; Areef Ishani,^1,7 MD, MS

*Corresponding author

Author affiliations

1. Minneapolis VA Healthcare System, Minneapolis, MN

2. Cooperative Studies Program Coordinating Center, VA Boston Healthcare System, Boston, MA

3. Department of Biostatistics, Boston University School of Public Health, Boston, MA

4. Pharmacy Benefits Management Services, Department of Veterans Affairs, VA Greater Los Angeles Healthcare System

Department of Medicine, VA Greater Los Angeles Healthcare System, Los Angeles, CA

Department of Medicine, David Geffen School of Medicine at UCLA, Los Angeles, CA

5. Michael E. DeBakey VA Medical Center, Baylor College of Medicine, Houston, TX

6. University of Tennessee Health Science Center, Memphis, TN

7. Department of Medicine, University of Minnesota, Minneapolis, MN

References

1. Concannon TW, Meissner P, Grunbaum JA, et al. A new taxonomy for stakeholder engagement in patient-centered outcomes research. J Gen Intern Med. 2012;27(8):985-991. doi:10.1007/s11606-012-2037-1

2. Baquet CR, Commiskey P, Daniel Mullins C, Mishra SI. Recruitment and participation in clinical trials: socio-demographic, rural/urban, and health care access predictors. Cancer Detect Prev. 2006;30(1):24-33. doi: 10.1016/j.cdp.2005.12.001. Epub 2006 Feb 21. PMID: 16495020; PMCID: PMC3276312.

3. Fogel DB. Factors associated with clinical trials that fail and opportunities for improving the likelihood of success: A review. Contemporary Clinical Trials Communications. 2018 Sep;11:156-164. DOI: 10.1016/j.conctc.2018.08.001. PMID: 30112460; PMCID: PMC6092479

4. Ngune I, Jiwa M, Dadich A, Lotriet J, Sriram D. Effective recruitment strategies in primary care research: a systematic review. Qual Prim Care. 2012;20(2):115-23. PMID: 22824564.

5. Bower P, Wilson S, Mathers N. Short report: how often do UK primary care trials face recruitment delays? Fam Pract. 2007 Dec;24(6):601-3. doi: 10.1093/fampra/cmm051. Epub 2007 Sep 13. PMID: 17872907.

6. Mahmud A, Zalay O, Springer A, Arts K, Eisenhauer E. Barriers to participation in clinical trials: a physician survey. Curr Oncol. 2018 Apr;25(2):119-125. doi: 10.3747/co.25.3857. Epub 2018 Apr 30. PMID: 29719427; PMCID: PMC5927782.

7. Ishani A, Leatherman SM, Woods P, Hau C, Klint A, Lew RA, Taylor AA, Glassman PA, Brophy MT, Fiore LD, Ferguson RE, Cushman WC. Design of a pragmatic clinical trial embedded in the Electronic Health Record: The VA's Diuretic Comparison Project. Contemp Clin Trials. 2022 May;116:106754. doi: 10.1016/j.cct.2022.106754. Epub 2022 Apr 4. PMID: 35390512.

8. Ishani A, Cushman WC, Leatherman SM, Lew RA, Woods P, Glassman PA, Taylor AA, Hau C, Klint A, Huang GD, Brophy MT, Fiore LD, Ferguson RE; Diuretic Comparison Project Writing Group. Chlorthalidone vs. Hydrochlorothiazide for Hypertension-Cardiovascular Events. N Engl J Med. 2022 Dec 29;387(26):2401-2410. doi: 10.1056/NEJMoa2212270. Epub 2022 Dec 14. PMID: 36516076.

9. DeJonckheere M, Vaughn LM. Semistructured interviewing in primary care research: a balance of relationship and rigour. Fam Med Community Health. 2019 Mar 8;7(2):e000057. doi: 10.1136/fmch-2018-000057. PMID: 32148704; PMCID: PMC6910737.

10. Jamshed S. Qualitative research method-interviewing and observation. J Basic Clin Pharm. 2014 Sep;5(4):87-8. doi: 10.4103/0976-0105.141942. PMID: 25316987; PMCID: PMC4194943.

11. Caulfield, J. How to do Thematic Analysis: Step-by-Step Guide & Examples [Internet]. Scribbr; 2019 September 6 [revised 2023 June 22]. https://www.scribbr.com/methodology/thematic-analysis/.

12.Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa

13. Leatherman, S. M., Beaudette-Zlatanova, B., Robben, G., Glassman, P. A., Woods, P., Ferguson, R. E., Cushman, W. C., & Ishani, A. (2025). Provider experiences with and attitudes about an embedded pragmatic clinical trial. BMC primary care, 26(1), 121. https://doi.org/10.1186/s12875-025-02799-w

14. Johnston, S., Liddy, C., Hogg, W. et al. Barriers and facilitators to recruitment of physicians and practices for primary care health services research at one centre. BMC Med Res Methodol 10, 109 (2010). https://doi.org/10.1186/1471-2288-10-109

15. Asch, S., Connor, S. E., Hamilton, E. G., & Fox, S. A. (2000). Problems in recruiting community-based physicians for health services research. Journal of general internal medicine, 15(8), 591–599. https://doi.org/10.1046/j.1525-1497.2000.02329.x

16. Guillemin M, McDougall R, Martin D, Hallowell N, Brookes A, Gillam L. Primary care physicians' views about gatekeeping in clinical research recruitment: A qualitative study. AJOB Empir Bioeth. 2017 Apr-Jun;8(2):99-105. doi: 10.1080/23294515.2017.1305007. Epub 2017 Mar 16. PMID: 28949839.

17.Topazian, R., Bollinger, J., Weinfurt, K. P., Dvoskin, R., Mathews, D., Brelsford, K., DeCamp, M., & Sugarman, J. (2016). Physicians' perspectives regarding pragmatic clinical trials. Journal of comparative effectiveness research, 5(5), 499–506. https://doi.org/10.2217/cer-2016-0024

18. Somkin CP, Altschuler A, Ackerson L, Geiger AM, Greene SM, Mouchawar J, Holup J, Fehrenbacher L, Nelson A, Glass A, Polikoff J, Tishler S, Schmidt C, Field T, Wagner E. Organizational barriers to physician participation in cancer clinical trials. Am J Manag Care. 2005 Jul;11(7):413-21. PMID: 16044978.