Use of an Electronic Patient Portal Among Disadvantaged Populations

Abstract

Electronic patient portals give patients access to information from their electronic health record and the ability to message their providers. These tools are becoming more widely used and are expected to promote patient engagement with health care.
To quantify portal usage and explore potential differences in adoption and use according to patients' socioeconomic and clinical characteristics in a network of federally qualified health centers serving New York City and neighboring counties.
Retrospective analysis of data from portal and electronic health records.
74,368 adult patients seen between April 2008 and April 2010.
Odds of receiving an access code to the portal, activating the account, and using the portal more than once
Over the 2 years of the study, 16% of patients (n = 11,903) received an access code. Of these, 60% (n = 7138) activated the account, and 49% (n = 5791) used the account two or more times. Patients with chronic conditions were more likely to receive an access code and to become repeat users. In addition, the odds of receiving an access code were significantly higher for whites, women, younger patients, English speakers, and the insured. The odds of repeat portal use, among those with activated accounts, increased with white race, English language, and private insurance or Medicaid compared to no insurance. Racial disparities were small but persisted in models that controlled for language, insurance, and health status.
We found good early rates of adoption and use of an electronic patient portal during the first 2 years of its deployment among a predominantly low-income population, especially among patients with chronic diseases. Disparities in access to and usage of the portal were evident but were smaller than those reported recently in other populations. Continued efforts will be needed to ensure that portals are usable for and used by disadvantaged groups so that all patients benefit equally from these technologies.

Full text

Use of an Electronic Patient Portal Among Disadvantaged
Populations
Jessica S. Ancker, MPH, PhD
1
, Yolanda Barrón, MS
1
, Maxine L. Rockoff, PhD
2
,
Diane Hauser, MPA
3
, Michelle Pichardo, MPH
3
, Adam Szerencsy, DO
3,4
, and Neil Calman, MD
3
1
Departments of Pediatrics and of Public Health, Weill Cornell Medical College, Division of Quality and Medical Informatics, New York, NY, USA;
2
Department of BiomedicalInformatics, Columbia University, New York, NY, USA;
3
Institutefor Family Health,New York, NY, USA;
4
New York University,
Langone Medical Center, New York, NY, USA.
BACKGROUND: Electronic patient portals give patients
access to information from their electronic health
record and the ability to message their providers. These
tools are becoming more widely used and are expected
to promote patient engagement with health care.
OBJECTIVE: To quantify portal usage and explore
potential differences in adoption and use according to
patients' socioeconomic and clinical characteristics in a
network of federally qualified health centers serving
New York City and neighboring counties.
DESIGN: Retrospective analysis of data from portal and
electronic health records.
PARTICIPANTS: 74,368 adult patients seen between
April 2008 and April 2010.
MAIN MEASURES: Odds of receiving an access code to
the portal, activating the account, and using the portal
more than once
KEY RESULTS: Over the 2 years of the study, 16% of
patients (n=11,903) received an access code. Of these,
60% (n=7138) activated the account, and 49% (n=5791)
used the account two or more times. Patients with chronic
conditions were more likely to receive an access code and
to become repeat users. In addition, the odds of receiving
an access code were significantly higher for whites,
women, younger patients, English speakers, and the
insured. The odds of repeat portal use, among those with
activated accounts, increased with white race, English
language, and private insurance or Medicaid compared to
no insurance. Racial disparities were small but persisted
in models that controlled for language, insurance, and
health status.
CONCLUSIONS: We found good early rates of adoption
and use of an electronic patient portal during the first
2 years of its deployment among a predominantly low-
income population, especially among patients with chron-
ic diseases. Disparities in access to and usage of the portal
were evident but were smaller than those reported
recently in other populations. Continued efforts will be
needed to ensure that portals are usable for and used by
disadvantaged groups so that all patients benefit equally
from these technologies.
KEY WORDS: personal health record; health information technology;
health disparities; chronic illness; insurance status.
J Gen Intern Med 26(10):1117–23
DOI: 10.1007/s11606-011-1749-y
© Society of General Internal Medicine 2011
INTRODUCTION
Improving patients' access to health information and their
communication with providers are important steps toward health
promotion, self-management of disease, and shared medical
decision-making. One technology that has the potential to
support these goals is the electronic patient portal, which gives
patients a view of selected data from a provider’s electronic health
record (EHR) and may allow patients to perform tasks such as
exchanging secure messages with physicians, requesting
appointments, and ordering medication refills. Because the
portal is dependent on a single EHR, some
1,2
have called it a
“tethered”personal health record (PHR).
Portals and PHRs are widely expected to help promote health
at both the individual and the population level by facilitating
patient–provider communication, helping people access and
manage personal health data, assisting in self-monitoring and
self-care, and engaging individuals more fully in their own
health care and wellness.
1–4
Portals and PHRs are currently
used by only about 7% of Americans
5
, but they are growing in
popularity
5,6
and are likely to become more common in light of
the Department of Health and Human Services' 2010 rule to
encourage "meaningful use" of health information technology,
under which providers and hospitals can receive incentives for
measures such as providing patients with electronic copies of
their health information upon request.
7
To benefit from such electronic communication, however,
consumers and patients will require computer and Internet
access
8
, computer literacy
9
, health literacy
10,11
,andnumer-
acy.
12
It is also possible that they will need particular attitudes,
such as confidence in their written communication skills or
trust in doctors or the health-care system.
13,14
Thus, it seems
probable that educationally, economically, or socially disadvan-
taged individuals will be less likely to use these electronic tools
than those who are better off, and that any benefits from portal
use will be unequally distributed.
Received September 29, 2010
Revised March 23, 2011
Accepted May 19, 2011
Published online June 7, 2011
1117

In fact, racial and educational disparities have been found in
use of portals.
15–17
Among portal registrants at the Cleveland
Clinic, whites were more likely than blacks to use the account
after registering for it.
15
A cohort study at Kaiser Permanente
Georgia found that portal registration was more likely among
whites, those with Internet access at baseline, and those with
more education.
16
At a public HIV/AIDS clinic, early adopters of
a customized patient portal were more likely to be white and
non-Hispanic than the overall clinic population.
17
Each of these studies, however, focused primarily on one part
of the process of engaging patients: either portal registration, or
portal usage among those already registered. If electronic portals
are to become useful across the entire population, disparities
should be minimized throughout the process of patient engage-
ment. Do disparities in usage arise because of unequal access to
portal accounts? If so, an appropriate response might be policies
to reduce disparities in portal access, such as automatically
providing access codes for all patients rather than assigning
them upon request of either the provider or the patient.
Alternately, do disparities arise when patients are given portal
access but do not become regular users? If this is the case, it
would suggest the need for measures to help patients to activate
these accounts and become regular users.
In this retrospective analysis, we examined demographic and
clinical factors associated with (a) receipt of a portal access code, (b)
activation of the portal account, and (c) use of the portal more than
once among patients in a safety-net clinic over 2 full years of data.
METHODS
Setting & Participants
We conducted our study using data from the Institute for Family
Health (IFH), a network of federally qualified health centers that
provides primary care to a predominantly low-income population
in New York City and the Hudson Valley north of the city, serving a
highly racially, ethnically, and socio-economically diverse popu-
lation. IFH uses the EpicCare EHR and in 2008 deployed Epic's
English-language MyChart patient portal. The portal went live in
April 2008; New York City patients began to be enrolled
immediately, and Hudson Valley patients 6 months later. In
February 2009, IFH launched voluntary provider training and a
poster campaign within all its health centers, but there was no
policy requiring systematic enrollment for all patients. An access
code for the portal could be generated by the provider upon
request of either the provider or the patient; after receiving an
access code, the patient had to visit the portal website to activate
the account, and then could use the portal freely.
To be included in the analysis, an individual had to be an
active patient of the Institute, defined as having had at least one
encounter at an IFH health center between when the patient
portal went live in April 2008 and April 2010. Patients under 18
were excluded.
Data were obtained from the Institute's EHR system and its
patient portal (EpicCare and MyChart, Epic Systems, Verona,
Wisconsin). The portal database was queried for patient-level
data about access codes, activation of portal accounts, and
usage. The patient's primary care provider, demographic char-
acteristics (sex, age, race, preferred language), and insurance
status were obtained from the EHR data. In addition, we
identified presence of nine chronic conditions by applying
algorithms developed by IFH for quality improvement purposes
to the ICD-9 codes in the patient's current problem list. These
algorithms scan the patient record for ICD-9 codes, and then
use the hierarchical structure of ICD-9 to group together all
codes pertaining to a specific chronic condition.
Statistical Analysis
Multivariate logistic regressions were used to determine char-
acteristics associated with receiving an access code, activating an
account, and using the account more than once. We controlled
for duration of portal access with two variables: number of
months since the access code was received, and the number of
months since the account was activated. Predictor variables
significant at the 0.05 level in bivariate analyses were included in
multivariate models; in all multivariate models, we also con-
trolled for demographic characteristics and duration of portal
access to address potential confounders. Because the individual
chronic disease diagnoses were components of the total number
of diagnoses, they could not be added to the same multivariate
models. Separate models were constructed for these, using
diagnoses significant on bivariate analysis, while also adjusting
for sex, age, race, language, insurance type, number of clinical
visits, health center location, number of months since access
code was issued, and the number of months since the account
was activated. Analyses were performed in SAS version 9.2 (SAS
Institute, Cary, NC).The study was approved by the IRBs of
Columbia University, Weill Cornell Medical College, and the
Institute for Family Health.
RESULTS
A total of 74,368 adults met the definition of active patients. They
were racially and ethnically heterogeneous, andonly 39% carried
commercial insurance (Table 1). Hypertension and hyperlipid-
emia were the most prevalent chronic conditions. Patients of the
Hudson Valley practices were older, more likely to have private
insurance or Medicare, and more likely to be white than patients
in the two urban boroughs. In the Bronx, nearly half of patients
were Hispanic, and 41% were covered by Medicaid.
Of the active patients, 11,903 (16%) received an access code
to the portal, 7138 (60%) of whom activated their accounts
("users") (Table 2). The account was used twice or more by 5791
“repeat users,”representing 81% of those who activated their
account and 49% of the total codes issued. Median frequency of
use was 0.8 times per month since account was activated. A
small number of patients logged in several times a day almost
every day, with the most frequent user accumulating 1532
logins over the two-year period. Of the 439 providers identified
as caring for eligible study patients, 289 (65.8%) had patients
who received a portal access code. These providers had a
median of eight patients receiving access codes (interquartile
range: 1—38).
Issuing of Access Codes
In bivariate analyses, the following variables were statistically
significantly associated with likelihood of getting an access code:
sex, age, race, language, insurance type, clinic location, number
of encounters, number of diagnoses, and the following individual
diagnoses: alcoholism, hepatitis, HIV/AIDS,asthma, depression,
1118 Ancker et al.: Use of an Electronic Patient Portal JGIM

and hyperlipidemia. Multivariate models (Table 3,column1)
showed that access codes were significantly more likely to be
issued to women, younger patients, whites, speakers of
English and languages other than English or Spanish, and
those with insurance, more clinical visits, and more chronic
illnesses. Patients in the urban sites were more likely to
receive access codes than patients in the Hudson Valley;
however, Hudson Valley sites implemented the portal
6 months after New York City sites. Stratified analyses (not
shown) demonstrated that all site-specific trends were similar
to the overall trends except with respect to age. In Manhattan,
odds of receiving a code was not associated with age, whereas in
the Bronx, older patients were less likely to receive a code and in
the Hudson Valley, older patients were more likely to receive a
code.
The racial disparities reflected by the odds ratios in Table 3,
although statistically significant, were relatively small in abso-
lute size. That is, in the complete data set, 16% (11,903/74,368)
of all patients received access codes; among black patients, the
proportion was 16% (1937/11,909); among white patients, it
was 18% (5887/32,638); and among Hispanic patients, it was
15% (2167/14,133).
Table 1. Characteristics of Institute for Family Health Patients by Health Center Location
Location of health center
Patient Characteristic Manhattan Bronx Hudson Valley All sites
N= 31,414 N= 15,667 N= 27,287 N= 74,368
Female Sex, n (%) 19,983 (64) 10,384 (66) 16,035 (59) 46,402 (62)
Mean age (SD) 37 (14) 39 (15) 45 (18) 40 (16)
Race/ethnicity, n (%)
Black/African American 4881 (16) 5103 (33) 1925 (7) 11,909 (16)
Hispanic 4857 (15) 7636 (49) 1640 (6) 14,133 (19)
White 14,358 (46) 399 (3) 17,881 (66) 32,638 (44)
Other 2868 (9) 606 (4) 631 (2) 4105 (6)
Missing/Unknown 4450 (14) 1923 (12) 5210 (19) 11,583 (15)
Preferred language, n (%)
English 28,619 (91) 10,789 (69) 23,584 (86) 62,992 (85)
Spanish 662 (2) 3499 (22) 678 (3) 4839 (6)
Other 2133 (7) 1379 (9) 3025 (11) 6537 (9)
Insurance type, n (%)
Private 15,091 (48) 3649 (23) 10,437 (38) 29,177 (39)
Medicaid 4321 (14) 6360 (41) 5993 (22) 16,674 (22)
Medicare 1556 (5) 1027 (6) 5948 (22) 8531 (12)
Other public* 820 (3) 1030 (7) 1150 (4) 3000 (4)
Uninsured 9626 (31) 3601 (23) 3759 (14) 16,986 (23)
Mean number of clinical visits over
2 years (SD) 5 (8) 7 (12) 7 (9) 6 (9)
Chronic conditions, n (%):
†
Hypertension 2796 (9) 2628 (17) 6275 (23) 11,699 (16)
Hyperlipidemia 2510 (8) 2464 (16) 4790 (18) 9764 (13)
Asthma 2146 (7) 1851 (12) 2420 (9) 6417 (9)
Diabetes 1223 (4) 1385 (9) 2688 (10) 5296 (7)
Depression 1906 (6) 1646 (11) 1050 (4) 4602 (6)
Drug abuse or dependency 588 (1.9) 493 (3.1) 662 (2.4) 1743 (2.3)
Chronic hepatitis (B, C, or other) 411 (1.3) 393 (2.5) 499 (1.8) 1303 (1.8)
Alcoholism 388 (1.2) 244 (1.6) 456 (1.7) 1088 (1.5)
HIV/AIDS 407 (1.3) 365 (2.3) 154 (0.6) 926 (1.3)
Mean number of chronic conditions (SD) 0.39 (0.82) 0.73 (1.09) 0.70 (0.99) 0.58 (0.96)
One or more chronic conditions, n (%) 7825 (25) 6584 (42) 11332 (42) 25741 (35)
Among those with one or more chronic
conditions, mean number of conditions (SD)
1.59 (0.89) 1.75 (1.03) 1.68 (0.86) 1.67 (0.92)
*Includes New York State programs Family Health Plus and Child Health Plus
†Conditions were not mutually exclusive, so percentages in this section do not sum to 100%
Table 2. Portal Access Among Institute for Family Health Patients by Health Center Location
Manhattan Bronx Hudson Valley Total
Number of access codes issued 7,856 1,907 2,140 11,903
Number of patients who activated
account (% of all patients)
4,340 (55%) 1,377 (72%) 1,421 (66%) 7,138 (60%)
Number of patients who activated
account the same day code was
issued (% of those who activated)
1,123 (26%) 929 (67%) 895 (63%) 2,947 (41%)
Number of patients who used portal
more than once (% of those who activated)
3,660 (84%) 1,040 (76%) 1,091 (77%) 5,791 (81%)
Median number of logins per month
(interquartile range)
0.9 (0.3, 2.1) 0.7 (0.2, 2.0) 0.8 (0.2, 2.0) 0.8 (0.3, 2.1)
1119Ancker et al.: Use of an Electronic Patient PortalJGIM

Activation of Account
In bivariate analyses, the following variables were significantly
associated with likelihood of activating the account: age, race,
insurance type, clinic location, number of encounters, number of
diagnoses, number of months since the access code was issued,
and the following three individual diagnoses: HIV, depression, and
hyperlipidemia. Multivariate models (Table 3,column2)showed
that activation was significantly more likely among older patients,
non-blacks, speakers of English or other languages compared to
Spanish speakers, the privately insured, patients in the Bronx,
and those with more clinical visits. Odds of activation increased
over time after the access code was issued, with 85% of patients
activating their code within 30 days. In the Bronx, odds of
activation were much higher than in the other regions. This may
have been at least in part because several Bronx clinicians helped
patients activate their account in the office on the same day the
access code was issued.
Repeat Portal Use
In analyzing likelihood of repeat use, we added "same-day
activation" as well as months since activating the account as
potential predictors (Table 3, column 3). In bivariate analyses,
the following variables were significantly associated with
repeat use: race, language, insurance type, clinic location,
number of encounters, number of diagnoses, number of
months since access code was issued, number of months
since account was activated, and 2 diagnoses: HIV and
hyperlipidemia. In the multivariate models (Table 3), we found
that repeat use was significantly more likely among whites,
English speakers, those with commercial insurance or Medic-
aid, those in either Manhattan or the Bronx, and those with
more chronic illnesses. Also, odds of repeat use rose over time
after the date of activation. In contrast, those who activated
their account the same day they received an access code were
less likely to become repeat users.
Among portal users, frequency of portal use was correlated
with number of clinical visits (r=0.31, p < 0.001), number of
diagnoses (r= 0.18,p < 0.001), and age (r=0.08, p < 0.001). In
addition, age was associated with number of clinical visits (r =
0.13, p < 0.001) and number of diagnoses (r=0.41, p < 0.001).
Clinical Diagnoses
Table 4shows the results for each diagnosis in multivariate
models in which individual diagnoses were substituted for
diagnosis count. In bivariate analyses, the following diagnoses
were associated with at least one of the outcomes of interest:
hyperlipidemia, asthma, depression, chronic hepatitis, alcohol-
ism, and HIV /AIDS (individual lists of significant predictors
appear in footnotes to Table 4). The multivariate model results
show that hyperlipidemia, asthma, and depression were associat-
ed with higher odds of having a portal access code generated; HIV/
AIDS was associated with higher odds but the association was not
Table 3. Characteristics Associated with Odds of Receiving a Portal Access Code, Activating the Portal, and Using It More Than Once
Adjusted Multivariate OR (95% CI)
Patient Characteristic Odds of receiving a
portal access code
Odds of activating
portal account
Odds of using portal
more than once
Female Sex 1.06 (1.01, 1.11) 1.07 (0.98, 1.15) 1.15 (1.01, 1.32)
Age (10-year increment) 0.97 (0.96, 0.99) 1.05 (1.01, 1.08) 0.99 (0.93, 1.04)
Race/ethnicity
Black/African American Reference Reference Reference
Hispanic 1.20 (1.12, 1.30) 1.35 (1.18, 1.55) 1.13 (0.92, 1.40)
White 1.60 (1.50, 1.71) 1.69 (1.50, 1.90) 1.54 (1.26, 1.87)
Other 1.21 (1.10, 1.34) 1.49 (1.25, 1.78) 1.27 (0.95, 1.71)
Missing/Unknown 1.06 (0.97, 1.17) 1.46 (1.23, 1.74) 1.85 (1.37, 2.49)
Language
Spanish Reference Reference Reference
English 2.80 (2.45, 3.20) 1.60 (1.24, 2.07) 1.72 (1.20, 2.46)
Other 2.18 (1.84, 2.59) 1.71 (1.23, 2.40) 1.39 (0.86, 2.27)
Insurance Type
Private 4.10 (3.84, 4.37) 1.71 (1.51, 1.94) 1.67 (1.36, 2.01)
Medicaid 2.19 (2.02, 2.37) 1.23 (1.08, 1.45) 1.47 (1.16, 1.87)
Medicare 1.88 (1.69, 2.09) 1.29 (1.06, 1.58) 1.32 (0.97, 1.81)
Other public* 3.16 (2.81, 3.56) 1.70 (1.37, 2.13) 1.12 (0.81, 1.55)
Uninsured Reference Reference Reference
Number of clinical visits 1.042 (1.039, 1.044) 1.012 (1.007, 1.018) 1.005 (0.99, 1.01)
Number of chronic conditions 1.15 (1.13, 1.18) 1.01 (0.96, 1.05) 1.15 (1.06, 1.24)
Health center location
Manhattan 5.23 (4.94, 5.54) 0.72 (0.64, 0.80) 1.69 (1.41, 2.02)
Bronx 2.63 (2.42, 2.85) 2.09 (1.77, 2.46) 1.41 (1.11, 1.79)
Hudson Valley Reference Reference Reference
Number of months since code issued (Not applicable) 1.04 (1.03, 1.05) (Not applicable)
Number of months since activation (Not applicable) (Not applicable) 1.06 (1.05, 1.07)
Account activated on the same
day access code was generated
(Not applicable) (Not applicable) 0.82 (0.72, 0.94)
*Includes New York State programs Family Health Plus and Child Health Plus
1120 Ancker et al.: Use of an Electronic Patient Portal JGIM

statistically significant. Only HIV/AIDS was associated with
increased likelihood of activating the account once the access code
was generated. Only hyperlipidemia was statistically significantly
associated with increased odds of using the portal repeatedly.
DISCUSSION
In this large retrospective study of portal use at a network of
federally qualified health centers, we found differences in electronic
portal access and use on the basis of race, ethnicity, sex, language,
insurance type, age, and health status. These disparities were
evident first in differences in the likelihood of receiving an access
code. Some of the disparities persisted across the likelihood of
activating the portal account and likelihood of using it more than
once. For example, racial and ethnic minorities, non-English
speakers, and people without commercial insurance were less
likely to receive access to the portal, and then there were additional
decreases in their likelihood of activating the portal account and of
using it. By contrast, men were less likely than women to receive
portal access codes, but there were no significant gender-based
disparities in activation or usage. Also, older patients were less
likely to receive portal access, but older patients who did receive
portal access were actually more likely to activate the account.
The disparities could be caused by patient self-selection,
selection by clinicians, or some combination of these factors.
However, even among patients who did receive access codes,
blacks and Hispanics were less likely than whites to activate the
account as well as to use it repeatedly. The racial and ethnic
disparities remained in models that controlled for insurance
status , language, age, and number of visits and of chronic
conditions. This suggests that the racial disparity cannot be
entirely attributable to selection on the part of clinicians but
instead reflects in part structural factors facing patients (such as
lack of computer access) or individual factors (such as low health
literacy). A policy to reduce disparities in PHR usage should thus
provide support for patients through all three stages; access to
the portal, activation of the portal account, and portal use.
However, in our data, activating the account on the same day the
access code was generated was associated with a lower likelihood of
becoming a repeat user. During the analysis, we learned that
several clinicians in the Bronx had been making concerted efforts to
provide access codes and activate portal accounts during the office
visit; our findings suggest that the patients so selected did not
uniformly become regular portal users. We also found that a
diagnosis of depression was associated with increased likelihood
of receiving a portal access code, but not with likelihood of activating
the account or becoming a repeat user; during the analysis, we
learned that one mental health clinic had been making concerted
efforts to provide portal access codes to patients.
The racial disparities appear smaller than those reported in
other recent studies, although direct comparisons are challenging
because each of these studies had different inclusion criteria and
methods, and examined a different stage in the process. A study of
patients participating in a cohort study at Kaiser Permanente
Georgia focused on access to the portal account. In this cohort,
30% of the blacks registered for a portal account compared to 42%
of the whites.
16
Hispanics were too rare to be analyzed. By
contrast, at IFH, portal access codes were created for 16% of IFH's
black patients, 18% of the white patients, and 15% of the Hispanic
patients.
Two studies have examined use of portal accounts. In one
among Cleveland Clinic patients who had registered for an
electronic portal, black patients constituted 6% of portal users
and 11% of registered nonusers; Hispanics were 2% of portal
users and 1% of registered nonusers.
15
In a small study at a
public HIV/AIDS clinic, nonwhites were 22% of users and 44% of
nonusers with e-mail addresses; Hispanics were 15% of users and
26% of nonusers with e-mail addresses.
17
By contrast, at IFH,
black patients constituted 15% of portal users and 18% of
nonusers with accounts; Hispanics constituted 19% of portal
users and 17% of nonusers with accounts.
Several explanations are possible for the smaller racial
disparities evident in our data. This study was conducted several
years after the Cleveland Clinic study, during which the “digital
divide”between whites and blacks narrowed nationwide
8
;how-
ever, our data collection period overlaps with that of the San
Francisco study, making a secular trend explanation less likely.
The New York metropolitan population may have had better
access to the Internet through public facilities such as libraries
and community technology centers than populations in other
regions. Finally, the Institute for Family Health has a strong
organizational commitment to the care of underserved popula-
tions, which may have created a culture that promoted equity.
Not surprisingly, we found that patients who visited health
centers more often were more likely to receive codes and activate
Table 4. Clinical Diagnosis Groups Associated with Odds of Receiving a Portal Access Code, Activating the Portal, and Using It More than
Once
Multivariate OR (95% CI)
Patient Characteristic Odds of receiving a
portal access code
1
Odds of activating
portal account
2
Odds of using portal
more than once
3
Presence of:
Hyperlipidemia 1.65 (1.54, 1.76) 1.11 (0.99, 1.25) 1.27 (1.06, 1.53)
Asthma 1.27 (1.18, 1.36) ––
Depression 1.13 (1.04, 1.23) 1.10 (0.96, 1.26) –
Chronic hepatitis (B, C, other) 0.96 (0.82, 1.14) ––
Alcoholism 0.92 (0.77, 1.10) ––
HIV/AIDS 1.17 (0.98, 1.39) 1.56 (1.19, 2.07) 1.15 (0.74, 1.78)
1. This model included only diagnoses statistically significant on bivariate analysis, which were: hyperlipidemia, asthma, depression, chronic hepatitis,
alcoholism, HIV/AIDS
2. This model included only diagnoses statistically significant on bivariate analysis, which were: hyperlipidemia, depression, HIV/AIDS
3. This model included only diagnoses statistically significant on bivariate analysis, which were: hyperlipidemia, HIV/AIDS
1121Ancker et al.: Use of an Electronic Patient PortalJGIM

their accounts; this relationship persisted even in models that
controlled for age and health status, as indicated by number of
chronic conditions. Age, number of clinical visits, number of
chronic conditions, and frequency of portal use were all
correlated with each other. Hyperlipidemia and HIV/AIDS were
the only chronic conditions associated with increased likelihood
of receiving portal access, activating the account, and becoming
a repeat user (although some of these associations missed
statistical significance). Surveys and qualitative research have
suggested that patients with chronic diseases have substantial
information needs that could be met through an electronic
portal, such as test results with explanatory information,
personal history data, and secure messaging.
5,18,19
Portals have
been used to screen patients for chronic conditions.
20
However,
Weingart et al. reported that enrollees of an Internet portal had
fewer medical problems than non-enrollees; enrollees were also
more affluent and younger than non-enrollees.
21
In the HIV
clinic study described above, portal users were as likely as
nonusers to have a diagnosis of AIDS, but were more likely to be
receiving antiretrovirals, to have undetectable viral load, and to
have high CD4 counts.
17
Limitations and Strengths
Previous reports of disparities in portal usage have focused
primarily on either portal registration
16,17
or upon use of the
portal.
15
Our study differs from these by examining disparities
in the likelihood of receiving a portal access code, activating
the account, and repeatedly using the portal. In addition,
previous reports have either been among insured patients
15,16
or among patients with a single chronic illness
17
; our study
looks at the entire patient population of a community health
center network. This study also extends previous work by
reporting results in a large and racially, ethnically, and
socioeconomically diverse population.
However, there are limits to what can be learned purely
from quantitative data extracted from the portal and the EHR.
Patient-reported data or geographic data could extend this
work by examining patient-reported outcomes such as percep-
tions and satisfaction, as well as other predictors including
education, literacy, employment status, neighborhood, and
other socioeconomic indicators. Portal use has been associated
with education level in a largely white population.
22
Patient–provider relationships are another important
factor that would be best studied through qualitative
studies or surveys. For example, it is not clear to what
extent patient use of a portal is intrinsically motivated and
what extent it is the result of encouragement by providers.
We have evidence that provider behavior varied, with only
65.8% of providers issuing codes to their patients, and a
great amount of additional variability in the number of
portal users per provider, with one quarter of providers
issuing only one access code, and another quarter issuing
38 or more. However, we have no information about other
provider factors including attitudes or beliefs, particularly
their commitment to engaging patients through the portal.
Formostpatients,providersarelikelytobetheoneswho
introduce them to the portal and interest them in its use,
and changing provider behavior may be an effective way to
increase patient engagement. On the other hand, some
qualitative research has linked patient interest in portals
with dissatisfaction with the patient–provider relationship.
18
In either case, provider attitude and behavior are likely to be
key factors for future research.
CONCLUSIONS
A safety net provider achieved good early rates of adoption and
usage for an electronic patient portal among a predominantly
low-income patient population. Racial and economic dispari-
ties were evident at all stages of access to the portal, activation
of portal accounts, and usage of accounts, but these dispa-
rities were smaller than those previously reported in other
populations. Patients with chronic conditions were more likely
to use the portal. In order to ensure that any benefits of patient
portals accrue equitably to all patients, continued efforts will
be needed to ensure that all patients receive access to portal
accounts and receive adequate support to activate and use
these accounts. These efforts might include: organizational
policies to encourage health care providers to use and promote
portals, to offer accounts to all patients, and to make
particular outreach efforts to low-income and minority
patients; public policies to increase access to computers and
the Internet; partnerships in which health-care organizations
help teach patients computer literacy skills; and interface
redesign to improve portal usability and accessibility. Research
in this domain should include qualitative and survey studies
about patients' and providers' reasons for both using portals
and not using them, as well as quantitative and mixed
methods approaches to assess the impact of policies intended
to reduce disparities in usage. Finally, researchers should
continue to assess the potential impacts of portal usage on
health outcomes, health-related knowledge, access to health
care, and the quality, safety, and costs of care.
Acknowledgments: This work was funded by HRSA grant 1
H2HIT086130101. A portion of the descriptive statistics was
presented at the annual symposium of the American Medical
Informatics Association, November 16, 2010, Washington DC.
During a portion of the study, Dr. Ancker was supported by NLM
training grant T15-LM007079.
Conflict of interest: None disclosed.
Corresponding Author: Jessica S. Ancker, MPH, PhD; Departments
of Pediatrics and of Public Health, Weill Cornell Medical College,
Division of Quality and Medical Informatics, 402 E. 67th St., LA-251,
New York, NY 10065, USA (e-mail: jsa7002@med.cornell.edu).
REFERENCES
1. Kaelber DC, Jha AK, Johnston D, Middleton B, Bates DW. A research
agenda for personal health records (PHRs). J Am Med Inform Assoc.
2008;15(6):729–36.
2. Kaelber DC, Shah S, Vincent A, Pan E, Hook JM, Johnston D, et al.
The Value of Personal Health Records. Charlestown, MA: Center for
Information Technology Leadership, Healthcare Information and Man-
agement System Society (HIMSS); 2008.
3. Nazi KM, Hogan TP, Wagner TH, McInnes DK, Smith BM, Haggstrom
D, et al. Embracing a health services research perspective on personal
health records: lessons learned from the VA My HealtheVet System. J
Gen Intern Med. 2009;25(Suppl 1):62–7.
4. The Markle Foundation. Connecting for Health: A Public-Private Collab-
orative: The Markle Foundation; 2003.
5. Undem T. Consumers and health information technology: A national
survey. Oakland, CA: California HealthCare Foundation; 2010.
1122 Ancker et al.: Use of an Electronic Patient Portal JGIM

6. Halamka JD, Mandl KD, Tang PC. Early experiences with personal
health records. J Am Med Informat Assoc. 2008;15(1):1–7.
7. Centers for Medicare and Medicaid Services. Medicare and Medicaid
Programs; Electronic Health Record Incentive Program; Proposed Rule.
In: Services DoHaH, editor.: Federal Register; 2010 (January 13).
8. Smith A. Home Broadband Adoption 2010. Washington, DC: Pe w
Internet & American Life Project; 2010.
9. Poynton T. Computer literacy across the lifespan: a review with
implications for educators. Comput Hum Behav. 2005;21:861–72.
10. Nielsen-Bohlman L, Panzer AM, Kindig DA, Editors, Committee on
Health Literacy. Health Literacy: A Prescription to End Confusion.
Washington, DC: Institute of Medicine. The National Academies Press.
2004.
11. Zarcadoolas C, Pleasant A, Greer DS. Understanding health literacy: an
expanded model. Health Promot Int. 2005;20(2):195–203.
12. Ancker JS, Kaufman DR. Rethinking health numeracy: A multidisci-
plinary literature review. J Am Med Informat Assoc. 2007;14(6):713–21.
13. Hesse B, Nelson D, Kreps G, Croyle R, Arora N, Rimer B, et al. Trust
and sources of health information: the impact of the Internet and its
implications for health care providers: findings from the first Health Informa-
tion National Trends Survey. Arch Intern Med. 2005;165(22):2618–24.
14. Jacobs E, Rolle I, Ferrans C, Whitaker E, Warnecke R. Understanding
African Americans’views of the trustworthiness of physicians. J Gen
Intern Med. 2006;21(6):642–7.
15. Miller H, Vandenbosch B, Ivanov D, Black P. Determinants of personal
health record use. J Healthc Inform Manag. 2007;21(3):44–8.
16. Roblin DW, Houston TK, Allison JJ, Joski PJ, Becker ER. Disparities
in use of a personal health record in a managed care organization. J Am
Med Inform Assoc. 2009;16(5):683–9.
17. Kahn JS, Hilton JF, Nunnery TV, Leasure S, Bryant KM, Hare CB, et
al. Personal health records in a public hospital: experience at the HIV/
AIDS clinic at San Francisco General Hospital. J Am Med Inform Assoc.
2010;17:224–8.
18. Zickmund SL, Hess R, Bryce CL, McTigue K, Olshansky E, Fitzgerald
K, et al. Interest in the use of computerized patient portals: Role of the
provider-patient relationship. J Gen Intern Med. 2007;23(Suppl 1):20–6.
19. Winkelman W, Leonard K, Rossos P. Patient-perceived usefulness of
online electronic medical records: Employing grounded theory in the
development of information and communication technologies for use by
patients living with chronic illness. J Am Med Informat Assoc.
2005;12:306–14.
20. Leveille SG, Huang A, Tsai SB, Weingart SN, Iezzoni LI. Screening for
chronic conditions using a patient Internet portal: recruitment for an
Internet-based primary care intervention. J Gen Intern Med. 2007;23
(4):472–5.
21. Weingart SN, Rind D, Tofias Z, Sands DZ. Who uses the patient
Internet portal? Journal of the Medical Informatics Association (Sympo-
sium Supplement): The PatientSite experience; 2005.
22. Hassol A, Walker JM, Kidder D, Rokita K, Young D, Pierdon S, et al.
Patient experiences and attitudes about access to a patient electronic
health care record and linked web messaging. J Am Med Informat Assoc.
2004;11(6):505–13.
1123Ancker et al.: Use of an Electronic Patient PortalJGIM