Patients with heart failure (HF) who are elderly, women, Black race, or Hispanic ethnicity are historically underrepresented in clinical trials. 1 In‐person recruitment is a barrier to participation that virtual enrollment may overcome. 2 Designed as a pragmatic trial with virtual enrollment (via email, text message, and telephone call) of patients seen in a HF clinic, the PRO‐HF (Patient‐Reported Outcome Measurement in Heart Failure Clinic) trial evaluates the impact of routine assessment of patient‐reported health status before HF clinic visits. 3 The effect of pragmatic, virtual trial design on representation is unknown. We sought to identify shared characteristics of enrolled PRO‐HF trial patients.
In the PRO‐HF trial, we recruited adults with Stanford HF clinic appointments between August 30, 2021 and June 30, 2022, via email 7 to 10 days previsit. Patients (n=372) were excluded because of competing trials with routine health status assessment. Consented patients completed baseline assessments via the Kansas City Cardiomyopathy Questionnaire‐12 online before their clinic visits. Patients who did not respond to 2 enrollment emails were contacted by telephone (telephone call followed by text message), repeated 3 to 5 days before their clinic visit. Enrolled patients were randomized to routine Kansas City Cardiomyopathy Questionnaire‐12 assessment or usual care. Detailed trial methods have been published previously. 4 Data for this study are available from the corresponding author only on reasonable request.
We obtained electronic health records of eligible patients from the Stanford Research Repository. Patient characteristics were defined by the date of first clinic visit during the recruitment period. Baseline demographics included age, sex, self‐reported race and ethnicity, comorbidities, and insurance. On the basis of home address 9‐digit zip code, we included the area deprivation index as a measure of neighborhood social risk. Clinical characteristics included a history of HF diagnosis, baseline comorbidities and medications, outpatient clinical encounters, hospitalizations in the prior year, and select laboratory measurements.
We compared characteristics between those who did and did not enroll and by method of enrollment: email only versus follow‐up telephone (call, text, or both as secondary analysis). We used standardized mean differences (SMDs) via Cohen d to evaluate the magnitude of differences between groups. 5 An SMD of 0.1 to 0.5 was considered small, and an SMD of >0.5 was considered medium to large. 5 We compared statistical significance with t‐tests for continuous variables and χ2 tests for categorical variables. We performed a multivariable logistic regression model to determine associations between enrollment and select characteristics: age, sex, race, ethnicity, and prior ouptatient clinic visits. The Stanford Institutional Review Board approved this study.
Of 5112 eligible patients, 1248 (24.4%) enrolled in the PRO‐HF trial (Table 1). This included 520 (41.7%) patients enrolled by email only and 728 (58.3%) enrolled by telephone. Enrolled patients were a median age of 63.8 years (interquartile range, 51.7–72.7 years), and 38.9% were women.
Table 1.
Baseline Characteristics of Patients Invited to Enroll in the PRO‐HF Trial
| Variable | Enrolled (N=1248) | Did not enroll (N=3864) | SMD (enrolled vs did not enroll) | Adjusted odds ratio (95% CI) of enrollment* | Email enrollment (N=520) | Telephone enrollment (N=728) | SMD (email vs telephone) |
|---|---|---|---|---|---|---|---|
| Demographics | |||||||
| Age, y | 63.8 (51.7–72.7) | 66.1 (52.1–75.8) | 0.08† | 0.99 (0.99–0.99) | 66.2 (55.2–74.3) | 61.9 (49.3–72.1) | 0.27† |
| Female sex | 38.9 (485) | 38.7 (1496) | 0.01 | 1.03 (0.90–1.18) | 38.3 (199) | 39.3 (286) | 0.02 |
| Interpreter preferred | 0.1 (1) | 10.6 (409) | 0.48† | 0.0 (0) | 0.1 (1) | 0.05 | |
| Race | 0.34† | 0.26† | |||||
| Asian | 12.1 (151) | 14.8 (570) | 0.56 (0.46–0.69) | 10.8 (56) | 13.0 (95) | ||
| Black | 4.9 (61) | 6.9 (268) | 0.54 (0.40–0.73) | 3.5 (18) | 5.9 (43) | ||
| Native American | 0.7 (9) | 0.3 (13) | 1.86 (0.77–4.50) | 0.2 (1) | 1.1 (8) | ||
| Unknown | 14.7 (183) | 24.9 (864) | 0.49 (0.39–0.61) | 12.7 (66) | 15.1 (117) | ||
| Pacific Islander | 1.5 (19) | 2.0 (78) | 0.47 (0.28–0.79) | 0.8 (4) | 2.1 (15) | ||
| White | 66.1 (825) | 51.0 (1971) | Reference | 72.1 (375) | 61.8 (450) | ||
| Ethnicity | 0.19† | 0.23† | |||||
| Hispanic or Latinx | 8.0 (100) | 13.2 (509) | 0.76 (0.58–0.99) | 4.6 (24) | 10.4 (76) | ||
| Non‐Hispanic | 87.7 (1095) | 81.2 (3139) | Reference | 91.3 (475) | 85.2 (620) | ||
| Unknown | 4.3 (53) | 5.6 (216) | 1.05 (0.73–1.51) | 4.1 (21) | 4.4 (32) | ||
| ADI national ranking | 4.0 (1.0–13.0) | 5.0 (2.0–18.0) | 0.05 | 3.0 (1.0–11.0) | 5.0 (2.0–15.0) | 0.16† | |
| Diagnoses | |||||||
| Atrial fibrillation | 35.0 (437) | 28.8 (1111) | 0.13† | 33.5 (174) | 36.1 (263) | 0.06 | |
| CAD | 40.1 (500) | 32.6 (1261) | 0.15† | 41.5 (216) | 39.0 (284) | 0.05 | |
| COPD | 14.1 (176) | 12.2 (473) | 0.06 | 13.8 (72) | 14.3 (104) | 0.01 | |
| Depression | 12.7 (159) | 9.2 (355) | 0.11† | 9.6 (50) | 15.0 (109) | 0.16† | |
| Diabetes | 18.8 (235) | 17.4 (671) | 0.04 | 16.3 (85) | 20.6 (150) | 0.11 | |
| HF or cardiomyopathy | 87.3 (1089) | 68.4 (2644) | 0.47† | 82.5 (429) | 90.7 (660) | 0.24† | |
| Hypertension | 52.7 (658) | 44.8 (1732) | 0.16† | 51.9 (270) | 53.3 (388) | 0.03 | |
| PAD | 38.5 (481) | 24.2 (937) | 0.31† | 40.4 (210) | 37.2 (271) | 0.06 | |
| Elixhauser comorbidity score | 4.0 (2.0–7.0) | 3.0 (2.0–6.0) | 0.16† | 4.0 (2.0–6.0) | 4.0 (2.0–7.0) | 0.16† | |
| Acute care services | |||||||
| ED visits in prior year | 13.5 (168) | 11.2 (431) | 0.07† | 12.3 (64) | 14.3 (104) | 0.06 | |
| Hospitalizations, prior 90 d | 9.9 (123) | 9.0 (347) | 0.03 | 7.3 (38) | 11.7 (85) | 0.15† | |
| HF hospitalizations, prior 90 d | 4.7 (59) | 4.7 (180) | 0.00 | 3.3 (17) | 5.8 (42) | 0.12† | |
| Any hospitalization in prior year | 20.2 (252) | 16.8 (649) | 0.09† | 16.9 (88) | 22.5 (164) | 0.14† | |
| HF hospitalizations in prior year | 9.2 (115) | 8.1 (312) | 0.04 | 7.5 (39) | 10.4 (76) | 0.10 | |
| Insurance | 0.11† | 0.26† | |||||
| Medicare | 47.2 (589) | 50.4 (1946) | 51.7 (269) | 44.0 (320) | |||
| Medicaid | 7.7 (96) | 9.5 (369) | 4.2 (22) | 10.2 (74) | |||
| Private | 37.7 (471) | 32.9 (1273) | 35.6 (185) | 39.3 (286) | |||
| Other | 7.4 (92) | 7.2 (276) | 8.4 (44) | 6.6 (48) | |||
| Outpatient encounters | |||||||
| New patients with HF | 22.9 (285) | 27.2 (1050) | 0.10† | 0.82 (0.76–0.86) | 21.3 (110) | 24.1 (175) | 0.07 |
| No. of prior HF clinic visits, prior year | 5.0 (1.0–11.0) | 2.0 (0.0–8.0) | 0.25† | 5.0 (1.0–12.0) | 5.0 (1.0–11.0) | 0.05 | |
| Stanford primary care | 65.1 (809) | 46.1 (1781) | 0.39† | 2.11 (1.84–2.42) | 68.0 (351) | 63.1 (458) | 0.10 |
| Echocardiogram, prior year | 58.6 (728) | 53.6 (2072) | 0.10† | 56.6 (292) | 60.1 (436) | 0.07 | |
| Left ventricular EF | 0.12† | 0.34† | |||||
| EF≤40% | 27.9 (348) | 17.3 (667) | 20.8 (108) | 33.0 (240) | |||
| EF >40% and <50% | 17.1 (213) | 10.7 (413) | 14.6 (76) | 18.8 (137) | |||
| EF≥50% | 54.8 (684) | 43.2 (1671) | 64.4 (335) | 47.9 (349) | |||
| Missing | 0.2 (3) | 28.8 (1113) | 0.2 (1) | 0.3 (2) | |||
| Vital signs and laboratory values | |||||||
| BMI, kg/m2 | 27.0 (24.0–31.4) | 26.4 (23.3–30.6) | 0.15† | 26.7 (23.5–30.0) | 27.5 (24.2–32.4) | 0.26† | |
| eGFR <30 mL/min per 1.73 m2 | 3.4 (42) | 5.8 (226) | 0.19† | 2.5 (13) | 4.0 (29) | 0.10 | |
| eGFR 30–44 mL/min per 1.73 m2 | 5.9 (74) | 6.6 (254) | 5.6 (29) | 6.2 (45) | |||
| eGFR 45–59 mL/min per 1.73 m2 | 13.3 (166) | 11.0 (424) | 13.8 (72) | 12.9 (94) | |||
| eGFR ≥60 mL/min per 1.73 m2 | 59.4 (741) | 48.9 (1889) | 60.0 (312) | 58.9 (429) | |||
| eGFR missing | 18.0 (225) | 27.7 (1071) | 18.1 (94) | 18.0 (131) | |||
| Medication therapies | |||||||
| ACEI/ARB/ARNI | 45.7 (567) | 34.8 (1346) | 0.22† | 39.5 (204) | 50.0 (363) | 0.21† | |
| β‐Blocker | 54.1 (672) | 39.7 (1535) | 0.29† | 48.6 (251) | 58.0 (421) | 0.19† | |
| Loop diuretics | 27.0 (335) | 28.0 (1083) | 0.02 | 22.5 (116) | 30.2 (219) | 0.18† | |
| MRA | 30.0 (372) | 22.7 (879) | 0.16† | 24.4 (126) | 33.9 (246) | 0.21† | |
| SGLT2i | 15.5 (192) | 10.0 (385) | 0.17† | 9.3 (48) | 19.8 (144) | 0.30† | |
Continuous variables are expressed as median with 95% CI. Other variables are given as percentage (number). Email recruitment preceded telephone recruitment. ACEI indicates angiotensin‐converting enzyme inhibitor; ADI, area deprivation index; ARB, angiotensin receptor blocker; ARNI, angiotensin receptor/neprilysin inhibitor; BMI, body mass index; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; ED, emergency department; EF, ejection fraction; eGFR, estimated glomerular filtration rate; HF, heart failure; MRA, mineralocorticoid receptor antagonist; PAD, peripheral artery disease; PRO‐HF, Patient‐Reported Outcome Measurement in Heart Failure Clinic; and SGLT2i, sodium‐glucose cotransporter‐2 inhibitor.
On the basis of multivariable logistic regression model with select patient characteristics: age, sex, race, ethnicity, new Stanford HF clinic patient, and prior Stanford primary care visit.
The SMDs when the differences between groups have a P<0.05 (based on t tests and χ2 analyses for continuous and categorical variables, respectively).
Enrolled and non‐enrolled patients were similar by age and sex but had differences across race (SMD=0.34) and ethnicity (SMD=0.19). Greater proportions of enrolled patients identified as White race (66.1% enrolled versus 51.0% nonenrolled) or non‐Hispanic ethnicity (87.7% versus 81.2%) compared with patients who identified as Asian race (12.1% versus 14.8%), Black race (4.9% versus 6.9%), or Hispanic or Latinx ethnicity (8.0% versus 13.2%). All non‐White (those who identified themselves from a race other than the White race [Asian, Black, Native American, and Pacific Islander]) patients enrolled in greater proportions by telephone than by email, with the largest difference among Hispanic or Latinx patients (10.4% versus 4.6%). Patients requiring an interpreter enrolled at lower rates (0.1% versus 10.6%; SMD=0.48). Enrolled and nonenrolled patients had similar area deprivation index (SMD=0.05).
Enrolled patients were more likely to have an existing HF/cardiomyopathy diagnosis and be prescribed HF medications. Enrolled patients had more HF clinic encounters in the prior year than nonenrollees (SMD=0.25) and were more likely to be seen by Stanford primary care (SMD=0.39). Enrolled patients had a slightly higher comorbidity burden, as defined by the Elixhauser score (SMD=0.16).
Patients enrolled via email were more likely to be older (SMD=0.27), White race (SMD=0.26), and non‐Hispanic ethnicity (SMD=0.23). Patients enrolled via email/telephone were more likely to have Medicaid (SMD=0.26) and higher area deprivation index scores (SMD=0.16).
In an adjusted analysis, Asian, Black, Pacific Islander, and Hispanic patients had lower odds of enrollment, whereas greater odds of enrollment were associated with prior Stanford primary care or HF clinic visits (Table 1). No significant association was found between sex and enrollment.
Virtual enrollment in the PRO‐HF trial varied by patient characteristics and enrollment method. Among the strongest predictors of enrollment were prior primary care and HF clinic visits. This may reflect how trusting patient‐physician relationships enhance patient perceptions of trial participation. 6 We identified racial and ethnic disparities similar to traditional trials. Compared with email only, follow‐up telephone recruitment enrolled greater proportions of younger patients, historically marginalized racial and ethnic groups, and Medicaid recipients.
This study has several limitations. First, multicollinearity across patient characteristics is likely. Second, race was unknown for 25% of patients who did not enroll. Third, independent effects of telephone versus email cannot be assessed. Finally, virtual enrollment may vary across trials with greater participant burden.
Virtual clinical trials enable efficient enrollment of large populations. Leveraging multiple recruitment modalities is important for achieving diverse representation in virtual trials that may lead to improved generalizability.
Sources of Funding
The PRO‐HF (Patient‐Reported Outcomes in Heart Failure Clinic) trial is supported by the National Heart, Lung, and Blood Institute (1K23HL151672‐01) and Stanford institutional funding. The data collection is supported by the National Institutes of Health (UL1 TR001085).
Disclosures
Dr Sandhu is supported by the National Heart, Lung, and Blood Institute (1K23HL151672‐03) and has consulting relationships with Lexicon Pharmaceuticals and Reprieve Cardiovascular. Dr Rodriguez reports consulting relationships with Healthpals, Novartis, NovoNordisk, and AstraZeneca outside the submitted work. Dr Kalwani reports stock options from Gordy Health and funding from the US Agency for Healthcare Research and Quality (T32 HS026128). Drs Azizi, Hernandez, and Wang were funded by American Heart Association Strategically Focused Research Network. The remaining authors have no disclosures to report.
This article was sent to Francoise A. Marvel, MD, Guest Editor, for review by expert referees, editorial decision, and final disposition.
For Sources of Funding and Disclosures, see page 4.
References
- 1. Tahhan AS, Vaduganathan M, Greene SJ, Fonarow G, Fiuzat M, Jessup M, Lindenfeld J, O'Connor CM, Butler J. Enrollment of older patients, women, and racial and ethnic minorities in contemporary heart failure clinical trials: a systematic review. JAMA Cardiol. 2018;3:1011–1019. doi: 10.1001/jamacardio.2018.2559 [DOI] [PubMed] [Google Scholar]
- 2. Martin SS, Ou FS, Newby LK, Sutton V, Adams P, Felker GM, Wang TY. Patient‐ and trial‐specific barriers to participation in cardiovascular randomized clinical trials. J Am Coll Cardiol. 2013;61:762–769. doi: 10.1016/j.jacc.2012.10.046 [DOI] [PubMed] [Google Scholar]
- 3. Sandhu AT, Zheng J, Kalwani N, Gupta A, Calma J, Skye M, Lan R, Yu B, Spertus J, Heidenreich P. Impact of patient‐reported outcome measurement in heart failure clinic on clinician health status assessment and patient experience: a sub‐study of the PRO‐HF trial. Circ Heart Fail. 2023;16:e010280. doi: 10.1161/CIRCHEARTFAILURE.122.010280 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Kalwani NM, Calma J, Varghese GM, Gupta A, Zheng J, Brown‐Johnson C, Amano A, Vilendrer S, Winget M, Asch S, et al. The patient‐reported outcome measurement in heart failure clinic trial: rationale and methods of the PRO‐HF trial. Am Heart J. 2023;255:137–146. doi: 10.1016/j.ahj.2022.10.081 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Routledge; 1988. [Google Scholar]
- 6. Houghton C, Dowling M, Meskell P, Hunter A, Gardner H, Conway A, Treweek S, Sutcliffe K, Noyes J, Devane D, et al. Factors that impact on recruitment to randomised trials in health care: a qualitative evidence synthesis. Cochrane Database Syst Rev. 2020;10:MR000045. doi: 10.1002/14651858.MR000045.pub2 [DOI] [PMC free article] [PubMed] [Google Scholar]