Telemedicine utilization in the ambulatory palliative care setting: Are there disparities?

Background:

Multiple professional organizations, including the National Comprehensive Cancer Network and the American Society of Clinical Oncology, advocate for the early integration of palliative care for patients with cancer to supplement the provision of disease-directed therapy.^1–4 Palliative care services have been shown to improve quality of life, symptom burden, and caregiver outcomes, as well as reduce healthcare costs.^5–9 Yet, despite evidence of the benefits, there is variability in utilization of palliative care services, with many patients being referred late in their disease trajectories or not at all.^10–15 Additionally, studies suggest significant racial and ethnic disparities in access to palliative care services.^16–24

Delays in initiation of palliative care for patients with cancer may be related, at least in part, to a shortage of palliative care physicians²⁵ and geographic variation in availability.^26,27 Telemedicine has been proposed as one way to improve access to palliative care for patients with cancer.²⁸ In many respects, telemedicine is well suited to this population given dynamic changes in symptom burden, such as worsening pain, that warrant frequent reassessment by clinicians as well as functional impairment that may make traveling from home to an outpatient clinic burdensome.²⁹

Despite the promise of telemedicine, however, the enduring digital divide raises concerns about whether the expansion of telemedicine services will exacerbate disparities in healthcare delivery.³⁰ Black and Hispanic patients as well as those living in poverty are less likely to have access to internet-enabled devices and broad-band coverage.³¹ Older adults and people living with serious illness, like cancer, are also disproportionately affected by the digital divide, resulting in limited internet use for healthcare.^32–36 Importantly, among older adults, race, ethnicity, and socioeconomic status have been independently associated with less internet use.³³

The COVID-19 pandemic accelerated the national expansion of telemedicine services. In particular, outpatient care shifted from in-person to telemedicine in order to abide by physical distancing regulations, keep patients and providers safe, and accommodate an increase in hospital volume. Although the US Centers for Medicare & Medicaid Services broadened reimbursement criteria and health systems rapidly expanded telemedicine services, these services did not make up for an overall decrease in outpatient care.³⁷ Furthermore, emerging evidence suggests that Black and Hispanic patients as well as those living in poverty were less likely to access care through telemedicine services during the peak of the pandemic.^38–40

Given the importance of palliative care for patients with cancer, and the pre-pandemic evidence of disparities in access to palliative care services,^16–24 we examined factors associated with utilization of telemedicine as compared to in-person visits by patients with cancer in the ambulatory palliative care setting. Examining telemedicine utilization, especially in the context of its recent growth, is crucial to ensuring equitable access to this novel mode of healthcare delivery for patients with cancer.

Methods:

Setting

The Mount Sinai Health System includes an NCI-designated cancer center and 8 ambulatory sites across New York City (NYC). Utilizing a single electronic medical record (EMR) that connects all locations, we collected data on all cancer patients seen in Supportive Oncology clinic by Palliative Medicine clinicians (physicians and advanced practice providers) with an in-person office visit or telemedicine visit during the COVID-19 pandemic from March 1, 2020 to December 30, 2020. Telemedicine was defined as a completed and billed audio-only or video visit. In our Supportive Oncology clinic, video visits were used preferentially, and all video visits occurred on the patient portal. Audio-only visits were primarily employed for patients with planned video visits that were unable to be completed due to technological barriers (e.g. lack of internet-enabled device, lack of internet access, and inability to access patient portal). Planned video visits that were ultimately converted to audio-only visits were identified by looking at visits that were scheduled as video and billed as audio-only.

Sociodemographic data, including race, ethnicity, gender, marital status, primary language spoken (English, Spanish, or other), New York City borough of residence, and insurance type were collected using a report generated from the EMR. All registration staff in the health system undergo training that includes instructions on obtaining demographic information such as race/ethnicity and preferred language. This training is then repeated annually. Staff are instructed to ask patients to select the appropriate option.

We collected information on rates of enrollment in the patient portal (MyChart). MyChart status included activated (i.e. patient has accessed the portal), inactivated (i.e. patient has declined portal participation or has never received an activation code), and pending activation (i.e. patient requested an activation code and never activated the portal). This study was considered a quality improvement initiative, and data was collected as part of routine care. Therefore, the study met criteria for exemption from Institutional Review Board approval at the Icahn School of Medicine at Mount Sinai.

Statistical Analysis

We assessed baseline characteristics of patients by visit type (telemedicine versus in-person) using descriptive statistics. Standardized mean differences were calculated across the characteristics, where differences less than 0.10 are considered minor. Subsequently, we used a generalized estimating equation (GEE) model, where each observation represented one visit, to assess the association between visit type and patient characteristics. As a result, patients did not have to be classified into one of the two modalities (in-person versus telemedicine). Instead, each visit was incorporated as its own modality with data clustered by patient to handle repeated patient visits. We used an exchangeable correlation structure to account for multiple visits per patient and within-subject correlation. Model assumptions require a high number of clusters and independence among observations in different clusters, both of which were met by our data. Additionally, GEE models provide robust standard error estimations and interpretable parameter values at the population level.⁴¹ Statistical reporting was done according to the SAMPL guidelines.⁴² All analyses were performed using R Version 4.0.3.

Results:

Sample Characteristics

Patient characteristics are shown in Table 1. A total of 491 patients and 1,783 visits were identified between March 1, 2020 and December 31, 2020, including 1,061 (60%) in-person visits and 722 (40%) telemedicine visits. Of the telemedicine visits, 614 (85%) were video visits, and 108 (15%) were audio-only visits. More than half (57, 53%) of the audio-only visits were originally scheduled as video visits and switched over at the time of the appointment due to technological barriers. The majority of patients was less than 65 years old (56%). While female patients accounted for 51% of in-person visits, they accounted for 55% of telemedicine visits. Our sample was racially diverse: among all visit types, White patients accounted for 32%, Black patients for 26%, and Asian patients for 11% of visits. Ethnicity was largely unknown (86%) with 4% of patients reporting their ethnicity as Hispanic, and 8% of patients reported their primary language as Spanish. Commercial insurance (23%), Medicaid (25%), and Medicare (42%) were the dominant insurance types, although 5% of patients did not have insurance documented in the EMR. In our sample, 55% were unmarried, and more than a quarter (27%) did not have an activated patient portal.

Table 1.

Characteristics of patients who had Supportive Oncology visits

	All Visits	In-Person	All Telemedicine	Video	Telephone	Standard Mean Differenceb
Number of patientsa	491	444	255	215	86
Age (< 65 years), n (%)	274 (56%)	252 (57%)	158 (62%)	136 (63%)	49 (57%)	0.17
Female, n (%)	256 (52%)	224 (51%)	140 (55%)	120 (56%)	47 (55%)	0.12
Race, n (%)c						0.34
White	158 (32%)	138 (31%)	89 (35%)	73 (34%)	32 (37%)
Black	129 (26%)	117 (26%)	71 (28%)	62 (29%)	24 (28%)
Asian	52 (11%)	48 (11%)	28 (11%)	24 (11%)	6 (7%)
Other	150 (31%)	139 (31%)	67 (26%)	56 (26%)	24 (28%)
Unknown	2 (<1%)	2 (0.5%)	0 (0%)	0 (0%)	0 (0%)
Ethnicity, n (%)c						0.45
Hispanic	21 (4%)	18 (4%)	8 (3%)	8 (4%)	0 (0%)
Not Hispanic	48 (10%)	43 (10%)	28 (11%)	27 (12%)	8 (9%)
Unknown	422 (86%)	383 (86%)	219 (86%)	180 (84%)	78 (91%)
Insurance, n (%)						0.45
Commercial	115 (23%)	107 (24%)	74 (29%)	66 (31%)	18 (21%)
Medicaid	122 (25%)	112 (25%)	68 (27%)	56 (26%)	23 (27%)
Medicare	208 (42%)	182 (41%)	91 (36%)	75 (35%)	36 (42%)
Self-pay/Unknown	22 (5%)	21 (5%)	2 (<1%)	2 (1%)	0 (0%)
Reported multiple	24 (5%)	22 (5%)	20 (8%)	16 (7%)	9 (10%)
Primary language, n (%)						0.37
English	417 (85%)	374 (84%)	227 (89%)	190 (88%)	80 (93%)
Spanish	41 (8%)	39 (9%)	11 (4%)	10 (5%)	3 (3.5%)
Other/Unknown	33 (7%)	31 (7%)	17 (7%)	15 (7%)	3 (3.5%)
Marital status, n (%)						0.22
Not married	269 (55%)	242 (55%)	138 (54%)	114 (53%)	52 (60%)
Married	208 (42%)	188 (42%)	112 (44%)	97 (45%)	31 (36%)
Unknown	14 (3%)	14 (3%)	5 (2%)	4 (2%)	3 (4%)
Patient Portal, n (%)						0.36
Activatedd	359 (73%)	319 (72%)	218 (86%)	189 (88%)	70 (81%)
Inactivatede	50 (10%)	47 (10%)	16 (6%)	13 (6%)	5 (6%)
Pending Activationf	82 (17%)	78 (18%)	21 (8%)	13 (6%)	11 (13%)
Borough, n (%)						0.53
Manhattan	186 (38%)	165 (37%)	91 (36%)	71 (33%)	34 (40%)
Brooklyn	82 (17%)	72 (16%)	44 (17%)	38 (18%)	15 (17%)
Queens	94 (19%)	91 (20%)	51 (20%)	45 (21%)	18 (21%)
Bronx	48 (10%)	43 (10%)	23 (9%)	20 (9%)	6 (7%)
Staten Island	13 (3%)	12 (3%)	6 (2%)	5 (2%)	1 (1%)
Outside of borough	68 (14%)	61 (14%)	40 (16%)	36 (17%)	12 (14%)

^aPatients who had at least one visit in a specific category were included in that category; thus, patients are included in more than one category if they had more than one type of visit. As a result, the numbers in the first row do not add up (categorization is not mutually exclusive among patients).

^bTo meet the assumption for three independent groups, data for the standard mean difference calculation were collapsed to mutually exclusive categories by selecting one random visit per patient. Pairwise comparisons were calculated across all three groups (in-person, video, audio-only), and values were then averaged to obtain the reported result.

^cRace/ethnicity were self-reported.

^dActivated indicates that patient has a portal that they have accessed.

^eInactivated indicates that patient has declined portal participation or has never received an activation code.

^fPending

Factors Associated with Telemedicine Utilization

The adjusted associations between patient characteristics and telehealth utilization are shown in Table 2. Female patients were significantly more likely to utilize telemedicine than male patients (OR 1.46; 95% CI 1.11–1.90). Spanish-speaking patients (OR 0.32; 95% CI 0.17–0.61), those without insurance (OR 0.28; 95% CI 0.15–0.52), and those without an activated patient portal (Inactivated: OR 0.46; 95% CI 0.26–0.82; Pending Activation: OR 0.29; 95% CI 0.18–0.48) were less likely to utilize telemedicine. Although Black patients were not less likely to utilize telemedicine, this odds ratio was trending toward statistical significance (OR 0.72, 95% CI 0.50–1.01).

Table 2.

Association of visit type and patient characteristics^a

	In-Person versus Telemedicine
	OR	95% CI
Race (Ref= White)
Black	0.72	0.50 – 1.01
Asian	0.88	0.53 – 1.46
Other	0.77	0.52 – 1.14
Gender (Ref = Male)
Female	1.46	1.11 – 1.90
Age (Ref < 65)
65+	0.93	0.64 – 1.34
Insurance (Ref = Commercial)
Medicaid	0.91	0.64 – 1.30
Medicare	0.85	0.57 – 1.28
Self-pay/Unknown	0.28	0.15 – 0.52
Primary Language (Ref = English)
Spanish	0.32	0.17 – 0.61
Other/Unknown	0.98	0.52 – 1.85
Marital Status (Ref = Not married)
Married	1.20	0.89 – 1.62
Unknown	0.69	0.31 – 1.52
Patient Portal (Ref = Activated)
Inactivated	0.46	0.26 – 0.82
Pending Activation	0.29	0.18 – 0.48

Activation indicates that a patient requested an activation code and never activated the portal.

^aThe analysis reflects individual visits (in-person visits n = 1,061; telemedicine visits n = 722) and not individual patients. We used a generalized estimating equation model, where each observation represented one visit, to assess the association between visit type and patient characteristics adjusting for all other covariates. Due to unknown race, 3 observations were excluded from the dataset when fitting the generalized estimating equation model. Furthermore, ethnicity was excluded from the adjusted analysis due to a high percentage of unknown, and borough of residence was excluded from the adjusted analysis as it may not accurately reflect distance from the clinic.

Discussion:

Our study reveals disparities in telemedicine utilization in the ambulatory palliative care setting for patients with cancer who are male, Spanish-speaking, uninsured, or do not have an activated patient portal. These findings suggest that the recent shift to telemedicine as a substitute for in-person visits may exacerbate existing disparities in access to disease-directed therapy, symptom management, and serious illness communication.

Similar to prior studies outside of palliative care, we found disparities in telemedicine utilization for Spanish-speaking patients in this setting.^39,43,44 Even after adjusting for patient portal activation, we showed that Spanish-speaking patients are less likely to access palliative care services via telemedicine. Clearly, activating the patient portal is not synonymous with being able to navigate it. Patient portals are designed with limited attention to literacy and usability, which creates greater barriers for patients with limited English proficiency than for the general population.⁴⁵ Furthermore, adding interpreters to telemedicine visits is not straightforward, posing an additional obstacle.⁴⁶ Seriously ill patients with limited English proficiency have worse outcomes in terms of pain management, communication at the end of life, and hospice enrollment.⁴⁷ Our study raises concern for widening disparities for this patient population in the context of telemedicine.

Although Black patients were not less likely to utilize telemedicine in our Supportive Oncology clinic, our findings trend toward statistical significance. Prior studies have shown limited telemedicine utilization among Black patients, and there are several hypothesized explanations for this disparity.^38,43 Due to structural racism in the healthcare system, some Black patients may be more concerned about privacy and confidentiality during telemedicine visits.⁴⁸ As a consequence, they may be less likely to choose telemedicine even if it is offered. Furthermore, implicit and explicit bias may contribute to disparities in telemedicine utilization for Black patients. Prior work has demonstrated differences in how primary care appointments are scheduled for Black patients, and there are unanswered questions about whether Black patients are offered telemedicine appointments to the same extent as White patients.⁴⁹ Importantly, when Black patients do have access to palliative care services via telemedicine, existing evidence suggests improvement in quality of life. Specifically, the ENABLE CHF-PC (Educate, Nurture, Advise, Before Life Ends Comprehensive Healthcare for Patients and Caregivers) study evaluated a 16-week palliative care telemedicine intervention for patients with heart failure across two hospitals in the Southeast US. This study demonstrated that pain intensity and pain interference with activities of daily living decreased among Black patients.⁵⁰ Given that Black patients with cancer experience delays in disease-directed therapy, less aggressive symptom management, and limited serious illness communication,^16,51 building the evidence base for palliative care telemedicine interventions for this patient population should be a priority. Furthermore, in order to narrow disparities, rather than exacerbate them, ensuring equitable access to existing telemedicine services will be crucial.

Our findings add to the emerging evidence that patient portal access predicts telemedicine utilization.³⁵ Although audio-only visits were reimbursed by private and public insurers during the pandemic, video visits conducted on the patient portal were far more common in our ambulatory palliative care clinic. Given the need to prescribe opioids and other controlled substances, palliative medicine clinicians were required to use video during many of their patient encounters. During the COVID-19 pandemic, the Drug Enforcement Administration waived several prescribing regulations, however continued to mandate in-person or video visits for new controlled substance prescriptions.⁵² Furthermore, it is likely that some clinicians felt more comfortable using video even when not prescribing controlled substances to improve engagement with their patients.⁵³ It has been suggested that video visits are superior in facilitating patient-provider communication than audio-only visits, although the evidence is inconclusive.^44,53

Our study highlights how for patients with serious illness like cancer, many of whom are older, video visits pose unique obstacles. The majority of audio-only visits (53%) occurred due to patient technological difficulties when attempting to engage in video visits. Learning new technologies can be cognitively challenging and functional impairment, such as loss of fine motor skills, further complicates access.^32,36,54 In addition, many health systems require that video visits occur on the patient portal to comply with HIPAA.⁵⁵ Although this requirement protects patient privacy and confidentiality, it has the unintended consequence of limiting the use of other technologies (e.g. Zoom, Facetime) with which patients may be more familiar. To bridge the digital divide, the benefits of video platforms that do not rely on patient portal access, in addition to audio-only visits, should be weighed against potential harms to the quality of patient care.^44,56,57

Limitations:

Our study has several limitations. We had poor documentation of ethnicity, with our data indicating that 4% of patients were Hispanic and 86% of patients had an unknown ethnicity. Hispanic patients represent 28% of the Mount Sinai Health System’s primary geographic catchment area as well at 13% of our cancer patient population. However, 8% of our patient population spoke Spanish as their primary language, which is a good proxy for Hispanic ethnicity. Nevertheless, our ability to describe this patient population was incomplete. Second, our study focused on one health system in NYC with a robust ambulatory palliative care practice for cancer patients, which may not be representative of other settings. Nevertheless, our health system includes community hospitals alongside an NCI-designated cancer center with a diverse patient population. Lastly, we were unable to compare characteristics of patients who used video versus audio-only visits due to limited sample size.

Conclusions and Future Directions:

Given the benefits of telemedicine for patients with cancer, many of whom have functional impairment making it difficult to leave home and rapidly changing symptoms requiring frequent reassessment, efforts to mitigate disparities are required. Ensuring patients have access to internet-enabled devices and broadband coverage are necessary but not sufficient next steps. Closing the digital divide will require technological assistance, especially for patients who do not speak English, have cognitive impairment, or do not have caregiver support.^32,58 Furthermore, future work should address what care should be delivered via telemedicine from the perspective of patients, caregivers, and clinicians and what is more appropriate for in-person visits. In the wake of the COVID-19 pandemic, if equity is kept at the forefront, we have an opportunity to better meet the palliative care needs of patients with cancer through telemedicine.