Patient Portal Use Among Patients with Vascular Anomalies and their Caregivers

Abstract

Background:

Patients with vascular anomalies (VAs) experience poor communication and have unmet information needs. Online patient portals could mitigate communication barriers and support communication interventions. However, these portals are often underutilized.

Procedure:

We retrospectively queried audit-log data from the Electronic Health Record (EHR) of a single large academic healthcare center for all patients seen by clinicians from a multidisciplinary specialist clinic with a diagnosed VA from 1/2020 to 1/2024. We connected audit-log data with patient demographics to examine how patients used the portal, and whether use varied by patient characteristics.

Results:

We queried portal usage for 315 patients with vascular anomalies, of whom 43% were children, 19% were adolescents, and 38% were adults. Approximately half of patients’ portals were logged into during the study period (51%, n=162). Of users who ever logged into the portal, the median number of logins per year were 35 (interquartile range 15 to 95). Multiple regression results show that portal access was higher for patients who are White, reside in a metropolitan area, and have lower Area Deprivation Index. Of users who ever logged into the portal, 77% viewed clinician notes, 90% viewed test results, and 71% engaged in messaging with a clinician at least once.

Conclusion:

Half of patients and caregivers never use the portal, and patients from less urban areas with higher deprivation are even less likely to use the portal. As portals become more integrated into patient care, these inequities in portal access could lead to inequities in health outcomes.

INTRODUCTION

Vascular anomalies (VAs) are incurable congenital disorders that cause pain, infection, bleeding, thrombosis, deformity, and disability.¹ VAs result from overgrowth of blood and lymphatic vessels, and can be associated with overgrowth of muscle, fat, and bone.^1–3 They can affect any part of the body and range in severity from mild to life-threatening. Patients with VAs have worse health-related quality of life than the general population due to pain,^4,5 decreased physical function,^6–8 worse mental health,^6,8 persistent uncertainty,^9,10 and psychosocial distress.^11–13 Families struggle to access clinicians with expertise in VAs, which can lead to delayed diagnosis, unnecessary procedures, frustration, and fear for patients and families.^14,15

High-quality communication is essential to optimize care for VA patients and to reduce burden on caregivers. We previously found that better information exchange was associated with better physical health, mental health, and ability to navigate the healthcare system for patients with VAs.^15–17 VA Caregivers described how availability of high-quality information was essential to supporting their ability to access needed healthcare for their child.^15–17 Many caregivers believed that asking questions and seeking information was central to their roles as advocates for their child.^14,16 However, caregivers and adult patients struggle to find understandable and trustworthy VA information, and many depended on social media and internet searches for VA information.^14,15 The rarity of disease and lack of trustworthy information creates persistent uncertainties that lead to confusion, frustration, and emotional distress.¹⁴ Efforts to enhance communication about VAs could improve caregiver knowledge and support better mental health, physical health, and ability to navigate the healthcare system.

Online patient portals represent widely available tools that could mitigate communication barriers and support communication interventions. As a result of the 21^st Century Cures Act, which required hospitals to provide patients with electronic access to their medical record,¹⁸ patient portal access has become nearly ubiquitous across healthcare settings.¹⁹ Portals offer several communication benefits. Adult cancer patients report that using portals and reading notes helps them to understand their diagnosis and treatment, communicate with oncologists, and engage with information.²⁰ Adult patients also report that reading clinical notes provides a sense of control over their health,^21–24 and improves adherence to treatment and follow-up plans.²¹ Parents of pediatric patients who read portal notes also reported an improved understanding of the rationale for tests and referrals, improved ability to complete tests and referrals, and no detrimental effects on their trust in doctors.²⁵ Portals could also help overcome communication barriers. Given the wide availability of online portals and the federal mandate for transparency,¹⁸ portals represent a powerful tool to mitigate barriers to communication in VA care.

Portals are underutilized in serious pediatric illnesses, and early data suggest disparities in which families access these portals. One pediatric oncology center found that 34% of families never activated a proxy online portal account.²⁶ Patients and caregivers with lower education level, lower income, or from minority racial or ethnic communities were less likely to use online portals in other studies.^26–29 Other barriers to portal use in adult medicine include preference for in-person communication, lack of perceived need for portals, and lack of comfort with computers.³⁰ No prior studies have evaluated portal use characteristics in VA populations. Furthermore, most prior studies have focused on whether patients or caregivers have accessed the portal, but almost no studies have characterized which features of the portal these families use. In this study, we developed and analyzed a novel dataset to examine whether patients with VA are accessing their portal, and if and how they used different portal features.

METHODS

Data.

We retrospectively queried audit-log data from the Epic Electronic Health Record (EHR) system of a single large academic healthcare center. We included all patients (pediatric and adult) with a diagnosed VA seen by clinicians from a multidisciplinary VA clinic between 1/2020 to 1/2024. Audit-log data captured the date and time of all patient portal activity including logging in, viewing notes and test results, and messaging. We linked audit-log data with data on patient demographics to examine how use varied by patient characteristics.

Data Processing.

All patient portal activity of interest associated with an eligible patient was collapsed to a patient-level dataset and analyzed at the patient level. Because prior studies have shown that parents often use their adolescent’s login credentials to access their portal,³¹ we examined combined proxy and adolescent usage data together (collectively referred to as “users”). We did not have demographic data for individual users; therefore, the demographic data presented in the study is specific to the patient whose portal was accessed.

Outcome Measures.

We were interested in four different patient portal activities: logging in for any reason, viewing clinician notes, viewing test results, and messaging with a clinician. For each activity, we calculated two measures: (1) whether the user engaged in the activity and (2) how often they engaged in the activity.

To measure whether the user engaged in the activity, we created a binary measure indicating whether they engaged in the activity at least once during the study period. To measure how often they engaged in the activity, we calculated the average number of engagements per year, starting from the first day they logged into the portal to the last day of our study period. For example, if a user logged into the portal for the first time on January 1, 2021 and logged in 20 more times between January 1, 2021 and January 1, 2024, their average number of yearly logins was calculated as 21 divided by 1,095 days multiplied by 365 days, or 7 times a year on average.

For our measures of logging in for any reason, we counted only the number of unique logins that occurred at least 60 minutes from a user’s most recent previous login to avoid double counting logins that were due to technical errors or connection issues. For our measures of viewing clinical notes, we limited clinical notes to those written by clinicians only. For our measures of messaging with a clinician, we limited messages to those exchanged with a clinician only. In other words, we excluded notes and messages written by or to other staff (such as registration) or auto-generated by the EHR system (such as reminders). Clinicians were defined as physicians, advanced practice providers, nurses, or medical assistants.

Patient Characteristics.

Data on patient characteristics were pulled from the EHR at the time that the study was conducted (January 2024). Based on prior literature, we examined patient characteristics that we hypothesized could be related to patient portal access characteristics including sex assigned at birth, race, ethnicity, language preference, and the patients age in years. Patients’ race was categorized using the 5 US census categories for race, Asian/Pacific Islander, American Indian/Alaskan Native, Black/African American, White, and Other. Ethnicity (i.e., Hispanic and non-Hispanic) was captured as a separate variable. Due to differences in access permissions by age, we categorized patient age into 3 groups: pediatric (ages 1-11), adolescent (ages 12-17), and adult (ages 18+).

The data also included patients’ street addresses, which we used to determine the Area Deprivation Index (ADI)³², rurality, distance from hospital, and access to the Internet. The ADI provides a neighborhood-level assessment of socioeconomic disadvantage based on factors such as income, education, employment, and housing. We were unable to calculate the area deprivation index for 2 patients who only had a PO box listed. Rurality was determined using the core-based statistical area for assessing rurality.³³ Based on their ZIP code, patients were grouped into one of four categories: metropolitan, micropolitan, small town, and rural areas. Distance from hospital was measured using Google Maps. Access to the Internet was estimated using the Data Infrastructure Score (DIS). DIS was developed by the American Telehealth Association, with the goal of representing the degree of local infrastructure that exists to support patients in accessing digital healthcare tools. This index ranges from 0-100 based on broadband availability, number of modalities for connection, broadband speed, and relative cost of data services for a family budget. Higher scores indicate that a higher prevalence of connectivity. For ease of interpretation, we categorized ADI, distance from hospital, and DIS into quartiles.

Statistical Analysis.

We first examined the proportion of users who ever logged into the portal and the frequency of logins by patient demographics. Then, we examined whether users viewed clinical notes, test results, or messaged with a clinician among the subset of user who ever logged into the patient portal, along with how often they engaged in these activities by patient demographics. Finally, we ran multiple logistic regression models using a subset of patient characteristics as independent variables, and whether users ever engaged in each activity as the dependent variable. For frequency of logins, we used a zero-inflated negative binomial model where our predictors for both parts of the model were the same patient characteristics. For the multiple regression models, we dropped distance from hospital and the DIS index due to collinearity (Supplemental Table S1) and collapsed the race variable into a binary “White” or “not White” variable and the statistical area variable into a binary “metropolitan” or “not metropolitan” variable due to small sample sizes. For all regression results, we present the average marginal effect (AME) of a change in the patient characteristic on the outcome of interest (e.g., increase in the probability of an associated user engaging in the activity or number of engagements per year).

In supplementary analysis, we examined when users were most likely to log into their portals on weekdays or weekends, and time of day (i.e., 12:00- 8:00 am, 8:01-2:59pm, and 3:00-11:59PM). We also examined how quickly patients viewed their test result after a test result was released.

RESULTS

Patient characteristics of our study population are presented in Table 1. Our study included 315 patients, of whom roughly half were female, 84% identified as White, 18% identified as Black or African American, and 2% identified as Asian or Pacific Islander, 3% identified as Hispanic, 3% had a recorded preferred language other than English, and 73% lived in metropolitan areas.

TABLE 1.

Patient Demographics

Patient Demographics	All Patients (n=315)
Sex
Male	135 (43%)
Female	180 (57%)
Race
Asian/Pacific Islander	7 (2%)
Black/African American	30 (10%)
White	266 (84%)
Other*	12 (4%)
Ethnicity
Hispanic	8 (3%)
Non-Hispanic	307 (97%)
Language Preference
English	307 (97%)
Non-English	8 (3%)
Patient Age in Years
1-11 (full access)	137 (43%)
12-17 (proxy access)	59 (19%)
18+ (no access)	119 (38%)
ADI National Percentile
4-41	79 (25%)
42-62	81 (26%)
63-81	75 (24%)
82-100	78 (25%)
Missing	2 (1%)
Statistical Area
Metropolitan	230 (73%)
Micropolitan	42 (13%)
Small Town	35 (11%)
Rural	8 (3%)
Distance from Hospital
1-21	80 (25%)
22-47	79 (25%)
48-128	80 (25%)
130-865	76 (24%)
DIS Index
18-56	82 (27%)
57-69	77 (25%)
70-80	75 (25%)
81-89	69 (23%)
Missing	12

Note:

^*Other includes other or missing; totals may not sum to 100% due to rounding

Portal Logins.

Just over half, or 51% (n=162), of patients’ portals were ever logged into during the study period. Of portals that were ever logged into, the median number of logins per year were 35 with an interquartile range of 15 and 95. Characteristics of patients whose portals were logged into, and frequency of logins, are presented in Table 2.

TABLE 2.

Patient Demographics by Patient Portal Use

	Ever Login (n=315)		Login Frequency per Year | Ever Logged In (n=162)
	Never Login n(%)	Ever Login n(%)	Median (p25, p75)
All Patients	153 (49%)	162 (51%)	35 (15, 95)
By Demographics
Sex assigned at birth
Male	71 (53%)	64 (47%)	33 (14, 100)
Female	82 (46%)	98 (54%)	37 (18, 91)
Race
Asian/Pacific Islander	3 (43%)	4 (57%)	35 (23, 66)
Black/African American	13 (43%)	17 (57%)	22 (15, 46)
White	128 (48%)	138 (52%)	38 (18, 102)
Other	9 (75%)	3 (25%)	3 (3, 22)
Ethnicity
Hispanic	4 (50%)	4 (50%)	25 (8, 112)
Non-Hispanic	149 (49%)	158 (51%)	35 (18, 95)
Preferred Language
English	145 (47%)	162 (53%)	35 (15, 95)
Non-English	8 (100%)	0 (0%)	0 (0,0)
Patient Age in Years
1-11	68 (50%)	69 (50%)	37 (20, 109)
12-17	37 (63%)	22 (37%)	38 (14, 102)
18+	48 (40%)	71 (60%)	33 (7, 81)
ADI National Percentile
4-41	31 (39%)	48 (61%)	34 (17, 93)
42-62	27 (33%)	54 (67%)	39 (23, 101)
63-81	47 (63%)	28 (37%)	28 (7, 57)
82-100	48 (62%)	30 (38%)	35 (14, 100)
Statistical Area
Metropolitan	96 (42%)	134 (58%)	36 (18, 99)
Micropolitan	32 (76%)	10 (24%)	73 (35, 115)
Small Town	20 (57%)	15 (43%)	23 (8, 40)
Rural	5 (63%)	3 (38%)	40 (3, 111)
Distance from Hospital
1-21	29 (36%)	51 (64%)	38 (19, 75)
22-47	31 (39%)	48 (61%)	31 (14, 88)
48-128	45 (56%)	35 (44%)	44 (15, 122)
130-865	48 (63%)	28 (27%)	27 (11, 44)
DIS Index
18-56	50 (61%)	32 (39%)	35 (14, 74)
57-69	43 (56%)	34 (44%)	40 (23, 115)
70-80	31 (41%)	44 (59%)	32 (17, 59)
81-89	21 (30%)	48 (70%)	35 (13, 102)

Note: p-values calculated using t-tests and ANOVA tests, rows may not sum to 100% due to rounding

Multiple regression results show that patients who are White, reside in a metropolitan area, and have lower ADI (i.e. lower deprivation) were more likely have their portal logged into. Patients with non-English language preference never logged into their portals. Adolescent portals also appeared to have lower likelihood of logging in compared to pediatric or adult portals, though this difference did not reach statistical significance (p=0.07). Patients who were White and who lived in a metropolitan area were also more likely to have a higher frequency of portal logins. For example, compared to patients who live in zip codes in the lowest quartile of the ADI (4-41), patients who live in third highest quartile of ADI (63-81) were 18 percentage points less likely to have their portals logged into (95% Confidence Interval (CI): −0.34,−0.01), and of those ever logged into the portal, the frequency of log-ins were lower by 14 fewer times a year on average (95% CI: −32, 3). Compared to patients who live in non-metropolitan areas, those who live in metropolitan areas were 15 percentage points more likely to log into the patient portal (95% CI: 0.01,0.29), and when they did log into the portal, logged in on average 23 more times a year (95% CI: 8, 38). See Table 3 for complete results.

TABLE 3.

Adjusted Marginal Effects of Patient Demographics on Portal Use

	Average Marginal Effect (AME) on Probability of Ever Login*		Average Marginal Effect (AME) on Number of Logins Per Year
	AME (95% CI)	P-Value	AME (95% CI)	P-Value
Sex assigned at birth (Ref: Male)
Female	0.00 (−0.11, 0.11)	0.97	5 (−10, 19)	0.52
Race (Ref: Not White)
White	0.05 (−0.11, 0.2)	0.55	19 (6, 32)	0.003
Ethnicity (Ref: Non-Hispanic)
Hispanic	−0.06 (−0.39, 0.27)	0.73	−7 (−44, 30)	0.71
Patient Age in Years (Ref: 1-11 Years)
12-17	−0.13 (−0.27, 0.01)	0.07	−4 (−25, 16)	0.67
18+	0.10 (−0.02, 0.22)	0.10	3 (−12, 19)	0.69
ADI National Percentile (Ref: 4-41)
42-62	0.07 (−0.08, 0.22)	0.35	10 (−10, 30)	0.32
63-81	−0.18 (−0.34, −0.01)	0.04	−14 (−32, 3)	0.10
82-100	−0.16 (−0.32, 0.01)	0.07	−3 (−24, 19)	0.80
Statistical Area (Ref: Non-Metropolitan)
Metropolitan	0.15 (0.01, 0.29)	0.04	23 (8, 38)	0.002
n	313		313

Note: AME=Average Marginal Effect, CI= Confidence Interval, results based on multivariate logistic regression model and zero-inflated negative binomial model;

^*2 patients were dropped from this model due to missing ADI

Portal Activities.

Of all portals that were accessed, 77% were accessed to view clinician notes at least once, 90% to view test results at least once, and 71% to message with a clinician at least once. The median number of times clinician notes were viewed was 2 per year (Interquartile range (IQR): 0,7), the median number of times test results were viewed was 12 (IQR: 4,33), and the median number of engagements with clinician messages was 3 (IQR: 0,9). Table 4 describes the frequency with which these portal activities were engaged. Exploratory multiple regression results show that portal users of female patient charts were 11 percentage points more likely to view test results than those of male patients (95% CI: 0.01,0.22), and portal users of adult patients were 7 percentage points more likely than those of pediatric patients to message with a clinician (95% CI: 0.02,0.11). (Table 5)

TABLE 4.

Patient Portal Activity per Year (n=162)

	Ever Used (n (%))			Times Used per Year (Median (p25, p75))
	View Clinician Notes	View Test Results	Messaging with Clinician	View Clinician Notes	View Test Results	Messaging with Clinician
All Patients who Ever Logged In (n=162)	124 (77%)	146 (90%)	115 (71%)	2 (0, 7)	12 (4, 33)	3 (0, 9)
By Demographics
Sex assigned at birth
Male (n=64)	47 (73%)	54 (84%)	44 (69%)	1 (0, 7)	10 (2, 33)	2 (0, 9)
Female (n=98)	77 (79%)	92 (94%)	72 (71%)	2 (0, 7)	15 (5, 35)	3 (0, 9)
Race
Asian/Pacific Islander (n=4)	3 (75%)	4 (100%)	4 (100%)	3 (1, 3)	18 (4, 47)	2 (1, 3)
Black/African American (n=17)	13 (76%)	17 (100%)	8 (47%)	2 (0, 3)	17 (3, 27)	0 (0, 8)
White (n=138)	108 (78%)	124 (90%)	102 (74%)	2 (0, 8)	12 (4, 35)	3 (0, 10)
Other (n=3)	0 (0%)	1 (33%)	1 (33%)	0 (0, 0)	0 (0, 1)	0 (0, 6)
Ethnicity
Hispanic (n=4)	8 (50%)	4 (100%)	2 (50%)	1 (0, 33)	31 (4, 103)	0 (0, 5)
Non-Hispanic (n=158)	122 (77%)	142 (90%)	113 (72%)	2 (0, 7)	12 (4, 33)	3 (0, 9)
Patient Age in Years
1-11 (n=69)	55 (79%)	62 (90%)	49 (71%)	2 (0, 9)	10 (3, 18)	3 (0, 10)
12-17 (n-22)	16 (73%)	21 (95%)	17 (77%)	3 (0, 6)	11 (5, 33)	1 (0, 8)
18+ (n=71)	53 (75%)	63 (89%)	49 (69%)	2 (0, 6)	19 (4, 48)	3 (0, 10)
ADI National Percentile*
4-41 (n=48)	38 (79%)	44 (92%)	37 (77%)	2 (0, 7)	17 (3, 52)	2 (0, 9)
42-62 (n=54)	45 (83%)	49 (91%)	40 (74%)	4 (1, 8)	12 (5, 27)	4 (0, 11)
63-81 (n=28)	19 (68%)	24 (86%)	17 (61%)	1 (0, 2)	10 (2, 16)	1 (0, 9)
82-100 (n=30)	21 (70%)	27 (90%)	20 (67%)	3 (0, 9)	17 (6, 35)	4 (0, 9)
Statistical Area
Metropolitan (n=134)	106 (79%)	121 (90%)	94 (70%)	2 (0, 8)	15 (4, 35)	3 (0, 9)
Micropolitan (n=10)	8 (80%)	9 (90%)	9 (90%)	3 (0, 11)	10 (4, 37)	12 (3, 14)
Small Town (n=15)	8 (53%)	14 (93%)	10 (67%)	0 (0, 4)	10 (1, 12)	2 (0, 8)
Rural (n=3)	2 (67%)	2 (67%)	2 (67%)	1 (0, 6)	7 (0, 180)	3 (0, 9)
Distance from Hospital
1-21 (n=51)	41 (80%)	47 (92%)	40 (78%)	2 (0, 11)	17 (3, 36)	4 (1, 9)
22-47 (n=48)	39 (82%)	45 (94%)	31 (65%)	2 (1, 5)	12 (4, 35)	1 (0, 5)
48-128 (n=35)	27 (77%)	30 (86%)	25 (71%)	6 (0, 11)	13 (4, 48)	5 (0, 14)
130-865 (n=28)	17 (61%)	24 (86%)	19 (68%)	1 (0, 4)	10 (1, 16)	4 (0, 8)
DIS Index**
18-56 (n=32)	25 (78%)	29 (91%)	22 (69%)	3 (0, 8)	11 (4, 23)	4 (0, 10)
57-69 (n=34)	23 (68%)	28 (82%)	23 (68%)	3 (0, 7)	12 (6, 27)	3 (0, 12)
70-80 (n=44)	36 (82%)	42 (95%)	29 (66%)	2 (0, 6)	12 (4, 32)	3 (0, 6)
81-89 (n=48)	37 (77%)	43 (90%)	37 (77%)	2 (1, 10)	15 (2, 48)	2 (0, 7)

Note: For Notes, Test Results and Messages, all times accessed were added up and then divided by the total number of months the patient has had access and then multiplied by 12;

^*denominator is 160 due to incalculable ADI for n=2,

^**denominator is 158 due to incalculable DIS index for n = 4

TABLE 5.

Adjusted Marginal Effects (AME) of Patient Demographics on Probability of Engaging in Portal Activity

	Viewing Clinical Notes		Viewing Test Results		Messaging with a Clinician
	AME (95% CI)	P-Value	AME (95% CI)	P-Value	AME (95% CI)	P-Value
Sex assigned at birth (Ref: Male)
Female	0.04 (−0.10, 0.18)	0.56	0.11 (0.01, 0.22)	0.03	−0.65 (−460.79, 459.5)	1.00
Race (Ref: Not White)
White	0.16 (−0.06, 0.38)	0.16	0.00 (−0.15, 0.14)	0.98	−0.22 (−141.32, 140.89)	1.00
Ethnicity (Ref: Non-Hispanic)
Hispanic	−0.3 (−0.8, 0.2)	0.24	N/A*		0.57 (−76.02, 77.17)	1.00
Patient Age in Years (Ref: 1-11 Years)
12-17	−0.07 (−0.27, 0.13)	0.51	0.06 (−0.05, 0.16)	0.28	0.05 (−24.69, 24.79)	1.00
18+	−0.08 (−0.22, 0.06)	0.29	−0.04 (−0.15, 0.07)	0.49	0.07 (0.02, 0.11)	0.004
ADI National Percentile (Ref: 4-41)
42-62	0.01 (−0.15, 0.18)	0.89	0.00 (−0.12, 0.11)	0.95	0.01 (0.01, 0.01)	<0.001
63-81	−0.08 (−0.29, 0.14)	0.49	−0.04 (−0.19, 0.12)	0.62	0.37 (−6.07, 6.82)	0.91
82-100	−0.02 (−0.23, 0.18)	0.82	0.00 (−0.14, 0.15)	0.97	0.05 (−38.04, 38.13)	1.00
Statistical Area (Ref: Non-Metropolitan)
Metropolitan	0.16 (−0.08, 0.4)	0.20	−0.02 (−0.15, 0.11)	0.78	0.10 (−4.58, 4.79)	0.97
n	160		156		160

Note: AME=Average Marginal Effect, CI= Confidence Interval, results based on multivariate logistic regression model,

^*Hispanic (n=4) patients dropped due to perfect prediction (i.e., all Hispanic patients viewed test results at least once)

Timing of Portal Access.

Patients most commonly accessed the portal on weekdays between the hours of 8 AM and 3 PM. (Supplemental Table S2) Data on time of accessing test results was missing for 45 patients. Of the 101 patients that had detailed test data, we found that, on average, 44% of patients viewed test results within 2 hours, 19% of patients viewed test results between 2-24 hours, 5% of patients viewed test results between 24-48 hours, 11% of patients viewed test results between 24-48 hours, 11% viewed results between 2-7 days, and 22% viewed results more than 7 days after test results were released.

DISCUSSION

In this study of patient portal use among patients with VAs, we found that nearly half of patient portals were never logged into. Patients and proxies who did access the portal demonstrated relatively frequent usage, with a median of 35 logins per year. Nearly 90% of portal users viewed their test results, 77% of portal users viewed clinician notes, and 71% engaged in portal messages with a clinician at least once. Similar to previous studies, we found signs of socio-economic disparities in portal access.^34–37 Patients with non-English language preference never accessed their portals in this study. Portals for patients from non-metropolitan areas with higher deprivation were less likely to ever be accessed for any reason, though these disparities were lessened when looking at differences in portal activities (e.g., viewing notes and test results) for patient portals that had ever been accessed. These findings could be evidence of a persistent digital divide between populations with access to technologies such as devices and faster Internet speeds and those without. Furthermore, these findings could indicate that signing up for the portal is the biggest barrier to portal use for these patients, and we should further characterize the modifiable and non-modifiable barriers to accessing the portal for these families, especially those from underserved populations such as patients with non-English language preference. These future studies could support the development of interventions to address modifiable barriers and strengthen advocacy efforts for structural and systemic changes to address the non-modifiable barriers.

We found that portals for adolescent patients were less frequently accessed than portals for pediatric or adult patients (37% compared to 50% and 60%, respectively). Yet, when they were accessed, users of adolescent portals viewed clinical notes and test results at comparable rates to those of pediatric and adult portals. These findings may be explained by increased barriers to adolescent portals that were created to address confidentiality concerns. Prior work suggests that caregivers of adolescents face several barriers to accessing proxy portals because of state laws and hospital policies.^38,39 Some healthcare systems completely shut down the portal during adolescence, and others severely limit the information that is made available to the adolescent proxy portal due to confidentiality concerns. Furthermore, parents and adolescents often must complete onerous processes to grant proxy access to the parent. These data provide further evidence that adolescent portal policies might be impeding parental access and use of the patient portal.

Patient portals are becoming increasingly integrated into patient care. As observed in this study, many patients and caregivers use the portal to review results, read notes, and send portal messages. Some healthcare systems are prioritizing portal-based communication as the primary means of communication.³⁹ Furthermore, emerging technologies could leverage the portal to improve communication and care for patients. For example, our team is currently developing a communication tool for VAs that is integrated into the portal. Lastly, researchers are even using the portal to recruit for clinical trials and other research studies. As portal functionality continues to expand and improve, patients who do not access the portal are at risk of being left behind. Future studies should aim to identify modifiable factors that are impeding portal access. Furthermore, clinicians and informatics administrators must be mindful of these access disparities and ensure that every patient can receive high-quality care, whether they use the portal or not.

The results of this study should be interpreted with several limitations in mind. First, the study took place in a single academic medical center with a single Epic instance. Results might not be generalizable to users of other EHR systems whose patient portal features may vary. Second, sample sizes for Asian/Pacific Islander and Hispanic patients, and patients with non-English language preference were small, limiting our ability to generalize results to the greater demographic group at large. Third, ADI and DIS were based on patient zip code or mailing address and not necessarily representative of the individual patient. Finally, though not a limitation, it is important to note that the study focuses on patients with VA, a rare congenital disease. Given the catchment area of the study site, we are confident that the study population represents a significant proportion of patients in the region with a diagnosed VA. However, results may not apply to other patient populations who may have different healthcare needs.

CONCLUSION

To our knowledge, this is the first study to examine audit-log data for patient portal use among patients with VAs. Our study shows that half of patients and caregivers never use the portal, and patients from less urban areas with higher deprivation are less likely to use the portal. As portals become more integrated into patient care, these inequities in portal access could lead to inequities in health outcomes. Future studies should further characterize the modifiable and non-modifiable factors that influence portal use, such as portal registration workflows, and examine the association between portal use and patient outcomes, such as adherence.

Abbreviations:

VA

Vascular Anomalies

DIS

Data Infrastructure Score