Approximately 6.7 million American adults experience heart failure [1], which is a leading cause of morbidity and mortality globally [2]. There are currently four classes of guideline-directed medical therapies (GDMTs) shown to improve outcomes for patients with heart failure, which include β-blockers (BBs), angiotensin-converting enzyme inhibitors/angiotensin receptor blockers (ACEI/ARBs) or angiotensin receptor neprilysin inhibitors (ARNIs), mineralocorticoid receptor antagonists (MRAs), and sodium glucose cotransporter 2 inhibitors (SGLT2Is) [3]. Adherence to these prescribed therapies has been associated with reduced cost, reduced heart failure–related morbidity and mortality, and improved quality of life for patients with heart failure [4-7].

Patients with heart failure have increasingly been using remote care as part of their treatment course [8]. A systemic review of randomized clinical trials indicated that the use of telemedicine in the management of heart failure appeared to lead to similar health outcomes as face-to-face delivery of care [9]. However, the majority of these studies were conducted in randomized controlled trials (RCTs). With the exponential growth in telemedicine visits for outpatient care, few studies have reported real-world evidence (eg, using data from electronic medical records) on the association between telemedicine visits and health outcomes among outpatients. Studies using electronic medical record data have shown that telemedicine has improved medication adherence among patients seen in an outpatient gastroenterology clinic [10], as well as an improvement in mean monthly tobacco treatment for inpatient counseling and an increase in outreach visits in the telehealth period compared with the pretelehealth period [11]. One study found that hospitalized patients with heart failure who received an outpatient follow-up either via telemedicine or in-person had a lower 30-day readmission rate than those who received no follow-up [12]. Telehealth has reduced wait times for appointments and may increase clinician visit frequency, which may help improve medication adherence [13]. However, to our knowledge, no study has examined the potential impact of the type of visits on medicine adherence among patients with heart failure using electronic medical records. The difference between an RCT and the study using real-world data with respect to adherence is that adherence in an RCT is enforced to ensure any lack of efficacy of the tested drug is not due to low adherence [14,15]. Therefore, our study aimed to examine the association between telemedicine visits versus in-person visits on medication adherence to heart failure GDMT.

A total of 9521 individuals with heart failure were included in this analysis, with 830 individuals in the telemedicine visits group and 8691 individuals in the in-person office visits group (Figure 1, Table 1). Overall, the mean age was 76.7 (SD 12.4) years. Most of the patients were White (n=6996, 73.5%), followed by Black (n=1060, 11.1%) or Asian (n=290, 3%). Over half of the patients were male (n=5383, 56.5%), and over half were married or living with partners (n=4914, 51.6%). Most patients’ health insurance was covered by Medicare (n=7163, 75.2%), followed by commercial insurance (n=1687, 17.7%) and Medicaid (n=639, 6.7%). Most of the patients had comorbid medical conditions including hypertension (n=7892, 82.9%), cardiac arrhythmias (n=5691, 59.8%), obesity (n=3473, 36.5%), valvular disease (n=3374, 35.4%), peripheral vascular disorders (n=3205, 33.7%), diabetes without complications (n=2824, 29.7%), diabetes with complications (n=2017, 21.2%), chronic pulmonary disease (n=2360, 24.8%), and chronic kidney disease (n=1953, 20.5%). The rate of prescription for each GDMT category included was as follows: BB (n=7803, 82%), ACEI/ARB (n=6167, 64.8%), ARNI (n=1421, 14.9%), MRA (n=2017, 21.2%), and SGLT2I (n=667, 7%).

Baseline characteristics of patients in telemedicine and in-person office visit groups are displayed in Table 1. Individuals in the telemedicine visits group were younger (72.6 vs 77.1 years; P<.001), with a higher proportion of people who were African American (110/830, 13.3% vs 950/8691, 10.9%) or Asian (27/830, 3.3% vs 263/8691, 3.0%; P=.04), preferred speaking English (747/830, 90% vs 7088/8691, 81.6%; P<.001), and were married or living with partners (473/830, 57% vs 4441/8691, 51.1%; P=.005) compared to those in the in-person visits group. However, a lower proportion of patients in the telemedicine group had Medicare insurance (515/830, 62% vs 6648/8691, 76.5%; P=.005) or were prescribed ACEI/ARB therapy (492/830, 59.3% vs 5675/8691, 65.3%; P<.001). Individuals in the telemedicine visit group had a higher number of hospitalizations in the past year (mean 0.334, SD 0.909 vs mean 0.211, SD 0.641; P<.001) and a lower number of outpatient visits in the past year (mean 3.67, SD 2.94 vs mean 4.01, SD 3.39; P<.001).

A histogram plot was used to assess the distribution of the continuous PDC (Figure 2). Overall, the average PDC was 0.81 (SD 0.286) and 71.3% (6793/9521) of patients had a PDC≥0.8 (Table 2). In the unadjusted model, the PDC between telemedicine visits and in-person office visits groups was not statistically different (mean 0.794, SD 0.294 vs mean 0.812, SD 0.285), with a rate ratio of 0.98 (95% CI 0.95-1.00; P=.09) (Table 3). The ratio of the PDC by types of visits remained similar after adjusting for demographic covariates including age, sex, race, marriage, language, and insurance (rate ratio 0.99, 95% CI 0.96-1.01; P=.34); demographics and comorbidity index (rate ratio 0.99, 95% CI 0.96-1.01; P=.34); demographics, comorbidity index, and health care usage in the past year, including the number of hospitalizations, number of outpatient visits, and number of primary care provider visits (rate ratio 0.99, 95% CI 0.96-1.02; P=.44); and demographics, comorbidity index, health care usage in the past year, and neighborhood SES index (rate ratio 0.99, 95% CI 0.97-1.02; P=.49).

Similarly, without adjusting covariates, there was no significant difference in the percent of PDC ≥0.8 between the telemedicine visits and in-person office visits groups (573/830, 69% vs 6220/8691, 71.6%), with the odds ratio for adherence of 0.89 (95% CI 0.76-1.03; P=.12). The odds ratio of medication adherence to GDMT by types of visits remained the same after adjusting for demographics (odds ratio 0.93, 95% CI 0.80-1.10; P=.40); demographics and comorbidity index (odds ratio 0.93, 95% CI 0.80-1.09; P=.39); demographics, comorbidity index, and health care usage in the past year (odds ratio 0.94, 95% CI 0.81-1.11; P=.48); and demographics, comorbidity index, health care usage in the past year, and neighborhood SES index (odds ratio 0.96, 95% CI 0.82-1.13; P=.65).

Using the electronic medical record data from a large academic health care system, our results indicate that patients with heart failure had similar medication adherence to GDMT between telemedicine and in-person office visits. Our findings are important because we provide real-world evidence that, for patients with heart failure, telemedicine can be an approach to outpatient visits, which may be an alternative way to maintain the same medication adherence as in-person visits. The randomized clinical trials summarized by a systematic review conducted prior to COVID-19 indicated that the use of telemedicine in the management of heart failure appears to lead to similar health outcomes as face-to-face or telephone delivery of care [9]. Our study, using real-world data from electronic medical records, shows that patients with heart failure have no differences in medication adherence between telemedicine and in-person office visits. Despite the exponential growth in telemedicine visits for outpatient care, limited studies have examined the effect of telemedicine visits on medication adherence among patients with heart failure using electronic medical records. One study found that hospitalized patients with heart failure, who received outpatient follow-up via telemedicine, had a lower 30-day readmission rate than those who received no follow-up [12]; however, this study did not examine medication adherence. For a different medical condition, one study using electronic medical record data indicated that telemedicine improved medication adherence among patients seen in an outpatient gastroenterology clinic [10].

Our results indicate that, compared with participants who had gone to in-person office visits, the participants in the telemedicine group were younger, more likely to be African American or Asian, preferred speaking English, were married or living with partners, and had lower rates of Medicare insurance. The data from the NYULH system are uniquely suited to explore the digital disparities in telemedicine, given its well-developed digital health infrastructure [24]. Previous studies reported that the proportion of young African American individuals accessing care through telemedicine increased after COVID-19 [24,25]. Our finding that more individuals who had only telemedicine visits preferred speaking English is consistent with a prior finding [25], which might be due to the fact that patient portals are only developed in English [26]. Similarly, our finding that a higher proportion of patients in the telemedicine group were married or living with a partner, compared to the in-person visit group, is consistent with prior findings [27]. Our study adds to the literature, showing that patients with heart failure who were younger adults, African American or Asian American, preferred speaking English, or were married or living with partners may benefit from telemedicine visits for medication adherence to GDMT.

The limitations of this study include unavailable variables related to digital literacy in the electronic health record dataset and the homogeneity of the patient population that was mostly White and had health insurance covered by Medicare, limiting generalization to other populations such as individuals with Medicaid or differing digital health literacy. Additionally, medication adherence was defined based on pharmacy fill data, which might not accurately reflect true medication adherence to GDMT, though the PDC is a commonly used measure for medication adherence [18-20]. Some people may not meet the criteria for GDMT, but adherence is still important if therapy is prescribed. Moreover, causal inference cannot be made due to the cross-sectional nature of this study. Furthermore, during most periods of the study, SGLT2Is had not been approved for heart failure; we also included patients, including those with heart failure with preserved ejection fraction, for whom some of these medications may not be part of GDMT. However, our study focuses on adherence to medications prescribed by a provider, and we presume adherence to prescribed medications is important regardless of the indications for prescribing. We admit that the way the primary outcome adherence is measured does not take into account the dosage or reaching targets for GDMT, and someone on the lowest dose of all therapies would get a perfect score. However, the strength of the study is that we provide real-world evidence for the application of telemedicine in clinical practice. Future research should examine telemedicine effects gathered from multiple health systems.

In summary, using the electronic medical record data from a large academic health care system, our study indicates that patients with heart failure have no differences in medication adherence between telemedicine and in-person office visits. Our study also indicates that patients who were younger, were African American or Asian, preferred speaking English, or were married or living with partners might particularly benefit from telemedicine visits. Our findings are important for clinical practice because we provide real-world evidence that, for patients with heart failure, telemedicine visits can be an approach for outpatient visits. As telemedicine is more convenient and avoids transportation issues, it may be an alternative way to maintain the same medication adherence to GDMT as in-person visits for patients with heart failure.