Keywords electronic health records - eligibility determination - decision support techniques
- patient portals - cancer early detection
Background and Significance
Background and Significance
Lung cancer is responsible for over 25% of all cancer deaths in the United States,
making it the leading cause of cancer death.[1 ] In 2011, the National Lung Screening Trial (NLST) reported that screening high-risk
patients with low-dose chest computed tomography decreased relative lung cancer mortality
by 20%.[2 ] In the years following, several national organizations including the American Cancer
Society, the American College of Chest Physicians, and the United States Preventive
Services Task Force have endorsed annual lung cancer screening (LCS) for high-risk
current and former smokers.[3 ]
[4 ]
[5 ]
While LCS decreases lung cancer mortality, it also introduces risks in the form of
false-positive test results that could lead to invasive procedures and complications.
For this reason, the Centers for Medicare & Medicaid Services (CMS) announced that
it will cover LCS for high-risk current and former smokers with the requirement that
all patients first have a shared decision-making discussion about the risks and benefits
of screening.[6 ]
[7 ] Notably, benefits and harms of screening vary substantially according to an individual's
risk for developing lung cancer.[8 ]
[9 ]
While widespread recommendations for LCS have existed since 2014,[5 ]
[7 ] only approximately 4% of eligible Americans have been screened.[10 ] Multiple barriers contribute to the limited uptake of LCS. First, many primary care
providers are unfamiliar with LCS guidelines.[11 ] Lewis et al performed a survey[12 ] of primary care providers and demonstrated only 47% of providers knew at least three
of six guideline components for LCS with 24% not knowing any guideline components.
Ersek et al[13 ] surveyed family physicians and found most had incorrect knowledge about which organizations
recommended screening with many physicians continuing to recommend chest X-rays for
LCS. Not surprisingly, more than 50% of family physicians reported making 1 or no
screening recommendations in the prior year. Second, the time demands of clinical
practice interfere with clinicians' ability to educate patients about this new screening
modality and engage in shared decision-making discussions.[14 ]
[15 ] Third, many patients are unaware of LCS or that they qualify.[16 ]
[17 ] Lastly, conducting the mandated in-person shared decision-making discussion is difficult
given scheduling and care access limitations and the complexities of the varying risks
and benefits of screening by individual risk factors.
New strategies are needed to address these provider, patient, and system barriers
to LCS. The growing popularity of patient portals and mobile technologies offer novel
opportunities for new interventions. Directly reaching out to patients to inform them
of LCS and empower them to make a screening decision can address patient knowledge
barriers and bypass many provider and system barriers. Patient portals provide a secure
Web site where patients can login to access their health information such as laboratory
results, upcoming appointments, medication information, and even medical notes. Additionally,
patients can use the portal for bidirectional electronic communication with their
medical providers. To receive the maximum payments from Medicare, federal “meaningful
use” criteria require practices to demonstrate their patients use a portal, strongly
incentivizing health systems to adopt and implement this technology.[18 ]
[19 ]
[20 ] Although data are limited, recent studies suggest that approximately 50% of patients
have an active patient portal account, and higher rates of adoption are expected in
the next few years.[21 ] Most portals also feature smartphone apps, a key feature since over 75% of American
adults now own a smartphone.[22 ]
[23 ]
[24 ] Furthermore, ownership of smartphones is as prevalent among underrepresented minority
groups, and even 67% of those with annual household incomes less than $30,000 or only
a high school education have a smartphone.[23 ]
[25 ]
Rather than relying on busy providers to identify patients who might qualify for LCS
and offer screening, health systems could query their electronic health records (EHRs)
to identify potential screening candidates and reach out to them via the patient portal.
Such an approach would extend the delivery of care beyond the confines of a traditional
office visit. However, whether patients would act on information sent via the patient
portal, or whether they would even read portal messages, is unknown.
Objective
We sought to determine the feasibility of a digital health strategy for LCS delivered
via the patient portal independent of a medical visit. Our intervention uses an algorithm
to identify screening candidates, engages these at-risk patients via the patient portal,
and provides them with personalized risk assessments to help them make a screening
decision and receive screening if desired.
Methods
Study Design and Setting
We conducted a single-arm pragmatic study in a large academic health system that included
4 hospitals and a network of over 70 community-based clinics located in 19 counties
in North Carolina. All practices share a common EHR, Epic. In 2017, over 88,000 patients
in the health system aged 55 to 77 years old had an activated patient portal account.
The health system's Lung Screening program has been awarded the “Lung Cancer Screening
Center of Excellence” designation by the Lung Cancer Alliance.[26 ] It houses a LCS clinic staffed by a nurse practitioner who meets with patients for
a shared decision-making visit, orders the LCS if appropriate, and follows up on test
results. Since the study was completed, the health system has opened a second LCS
program in one of its affiliated hospitals. This pragmatic study was approved by the
Wake Forest Health Sciences Institutional Review Board (IRB00036974) with a waiver
of signed informed consent given that the interactive Web site delivers guideline-recommended
information consistent with usual care.
EHR Algorithm to Identify Screening Candidates
We developed an EHR algorithm to identify potential candidates for LCS based on age
and smoking history. Because the accuracy of recorded smoking histories was uncertain,
we only excluded patients who were recorded as being “never smokers” to capture as
many potential candidates as possible. We then excluded patients who may not be appropriate
for screening based on a prior history of lung cancer or those with a life-limiting
comorbidity (receiving treatment for cancer within the last 12 months; taking a medication
for dementia; or evidence of end-stage renal disease, cirrhosis, and congestive heart
failure).
mPATH-Lung Interactive Web Site
Interactive Web site design : Members of our team have previously developed and tested Web- and tablet-based decision
aids for colorectal cancer screening.[27 ]
[28 ]
[29 ] Based on this user-tested design, we created a Web-based decision aid for LCS, called
mPATH-Lung (mobile Patient Technology for Health - Lung). Notably, mPATH-Lung did
not replace shared decision making, but instead was designed to ensure patients were
well informed of risks and benefits prior to in-person shared decision making with
a medical provider. The interactive Web site is fully Web-based, requiring no download,
and is functionally available on both desktop and mobile browsers (mpath.wakehealth.edu/lung).
To keep the user interface simple and scalable for mobile devices, each screen displays
a single question with large intuitive response buttons or free-text entry blanks
([Fig. 1 ]). Users first answer 8 questions to determine their eligibility for LCS based on
the CMS criteria.[30 ] If users are eligible for screening, the decision aid shows them a brief video reviewing
the general benefits and risks of LCS that are mandated by CMS.[7 ] Immediately after the video, the decision aid allows users to estimate their personal
risks and benefits of LCS by answering 8 additional demographic and medical history
items. The interactive Web site then uses a model we developed[31 ] to determine and display the personalized risk–benefit screening information in
an infographic format ([Fig. 2 ]). This model uses the validated Prostate, Lung, Colorectal, and Ovarian Cancer Screening
Trial model of 2012[32 ]
[33 ] to calculate lung cancer risk and a reanalysis of the NLST results to estimate harms
and benefits. Participants not answering the additional optional items are shown the
risks and benefits of screening for the average-risk participant in the NLST.[2 ]
[31 ]
[34 ]
[35 ]
[36 ] The program closes by asking users if they would like to receive screening, and
it gives users who respond “yes” or “maybe” the telephone number for the LCS clinic,
staffed by a nurse practitioner, to schedule an appointment if they would like screening
or more information. Prior to screening, patients meet with the nurse practitioner
for an in-person shared decision-making visit.
Fig. 1 Example screenshot of mPATH-Lung patient-reported information.
Fig. 2 Example screenshot of mPATH-Lung personalized risk–benefit information.
Study Sample
We ran the EHR algorithm weekly to identify potential screening candidates who were
scheduled to see a Wake Forest primary care provider in the next 4 weeks and had logged
into their patient portal account within the last 90 days. A research assistant then
sent these patients an electronic message via the patient portal that displayed in
their inbox as being sent from the “mPATH Team.” No other information was visible
to the patient until the message was opened. Once opened, the message stated, “A new
screening test can significantly lower a person's chance of dying from lung cancer.
An automated analysis of your Wake Forest Baptist Health record indicates you may
qualify for this test. To learn more, click <hyperlink>” where the hyperlink contained
a unique study identifier and took the user directly to the mPATH-Lung landing page.
The message did not reference the patient's primary care practice or provider.
Outcome Measures
Our primary outcome of interest was the proportion of screening eligible participants
who completed LCS within 4 months (120 days) of the sent portal invitation. Secondary
outcomes included the reach of this digital-based approach to screening (defined as
the proportion of invited participants who visited and completed the mPATH-Lung eligibility
items), the proportion who read the portal message (defined as opening the message
in the patient portal) as captured by the EHR logs, screening decisions made, and
screening clinic appointments scheduled. A visit to mPATH-Lung was defined as clicking
on the hyperlink in the patient portal message. We defined completion of the interactive
Web site as answering all LCS eligibility questions, at which point mPATH-Lung either
informs patients they are ineligible for screening, or if eligible, displays the LCS
decision aid video. The Web site recorded time of visit and answers to all questions,
and we determined which participants visited the Web site via the unique study ID
embedded within each user's hyperlink.
Statistical Analysis
Our target sample size of 1,000 participants was chosen to provide a narrow 95% confidence
interval (±3%) around our primary and secondary outcomes. Chi-square tests were used
to assess differences in patient characteristics between those who did and did not
read the portal message and between those who did and did not visit the interactive
Web site. Statistical analyses were completed using SAS version 9.4. All analyses
were based on intention to treat.
Results
Between November 2016 and February 2017, we sent 1,000 patient portal invitations
to study participants. Of the messages sent, three were never received by the patient
due to technical errors but were retained in our data set per intention-to-treat principle.
Mean age was 65 years; 49% were male. Baseline characteristics of the study participants
are noted in [Table 1 ].
Table 1
Baseline demographic characteristics of study sample
Characteristic
Total sample
Read portal message[a ]
Visited interactive Web site[a ]
% (n )
% (n )
p -Value
% (n )
p -Value
Total
100 (1,000)
86 (862/1,000)
41 (404)
Age (y)
NS
NS
55–59
27 (272)
83 (225)
35 (94)
60–64
24 (239)
86 (206)
42 (100)
65–69
25 (247)
88 (217)
44 (110)
70–77
24 (239)
90 (214)
42 (100)
Race/Ethnicity
< 0.05
NS
Non-Hispanic White
84 (840)
88 (742)
41 (343)
Non-Hispanic Black
14 (137)
74 (101)
40 (54)
Hispanic
1 (9)
100 (9)
44 (4)
Others
1 (11)
91 (10)
27 (3)
Gender
NS
< 0.05
Male
49 (490)
87 (428)
33 (163)
Female
51 (507)
86 (434)
48 (241)
Insurance status
NS
NS
Commercial
45 (451)
86 (388)
39 (174)
Medicare
51 (512)
87 (447)
43 (218)
Medicaid
2 (19)
84 (16)
37 (7)
Uninsured
2 (15)
73 (11)
28 (5)
Smoking status
NS
NS
Current
19 (186)
85 (158)
44 (82)
Former
81 (811)
87 (704)
40 (322)
Abbreviation: NS, nonsignificant.
a Data abstracted at 120 days.
Messages Read and mPATH-Lung Visits
Out of 1,000 messages sent, 862 (86%) were read by participants. A total of 40% (404/1,000)
visited the interactive Web site ([Fig. 3 ]). A similar percentage read the messages regardless of age, gender, insurance status,
or smoking status ([Table 1 ]). The likelihood of reading the message varied by race/ethnicity, but a similar
proportion visited the interactive Web site regardless of race/ethnicity. Only gender
was associated with mPATH-Lung visits, with women being more likely to visit the interactive
Web site than men (48% vs. 33%, p < 0.05).
Fig. 3 Pragmatic trial flow diagram.
[Fig. 4 ] illustrates the percentage of the study sample who read the message and who visited
mPATH-Lung over time (from 0 to 120 days). The number of days to read the portal message
(of those who read the message) ranged from 0 to 249 with a median of 0.7 days (or
17 hours). Of the 404 participants who visited the interactive Web site, the number
of days from reading the portal message to visiting the app ranged from 0 to 75 with
a median of 0.4 days (9.6 hours). Among participants who visited mPATH-Lung, almost
all (96%) visited it within 1 day of reading the message.
Fig. 4 Time to event plot showing % of sample who read the message and % of sample who visited
the Web app over time (from 0 to 120 days).
Completion of mPATH-Lung
Of the 404 patients who visited the interactive Web site, 86% (349/404) completed
the core components. Ninety-nine of the completers were eligible for screening. Of
the 250 participants ineligible for screening, 13 (5%) were nonsmokers, 75 (30%) had < 30
pack years, and 162 (65%) quit smoking > 15 years ago.
Screening Decisions
A total of 81 screening eligible participants indicated a screening decision (24 selected
“Yes,” 36 selected “Maybe,” and 21 selected “No”). In a post hoc analysis, an individual's
risk of developing lung cancer was not associated with a decision to receive screening.
The median (range) risk for developing lung cancer for participants who wanted screening
was 4.2% (1.5–13.7%), for those who stated “Maybe” was 3.5% (0.4–21.5%), and for those
who stated “No” was 2.1% (0.6–13.4%), p = 0.09.
LCS Orders and Completion
[Table 2 ] depicts chest CT orders and completion stratified by study participants who did
not read the patient portal invitations, read the invitation but did not visit the
mPATH-Lung, and those who visited the interactive Web site. Only 1 to 5% of patients
completed an LCS clinic visit or completed any chest CT scan, and these proportions
did not differ whether patients read the portal message or visited mPATH-Lung. More
patients completed an LCS than those who visited the LCS clinic because primary care
providers could complete shared decision making and order screening independent of
the LCS clinic. Five of the 24 patients (21%) who expressed a desire to be screened
had a chest CT completed ([Table 3 ]).
Table 2
Chest CT orders and completion[a ]
Screening outcome
Did not read message
(n = 135)
Read message, did not visit mPATH-Lung
(n = 458)
Visited mPATH-Lung
(n = 404)
LCS clinic visit scheduled
1 (1%)
3 (1%)
4 (1%)
LCS clinic visit completed
1 (1%)
3 (1%)
3 (1%)
Any chest CT completed
5 (4%)
14 (3%)
22 (5%)
Screening or nodule follow-up chest CT completed
1 (1%)
6 (1%)
9 (2%)
Other chest CT completed
4 (3%)
8 (2%)
13 (3%)
Abbreviations: CT, computed tomography; LCS, lung cancer screening.
a All p -values > 0.05.
Table 3
LCS screening outcome compared with LCS interest
Screening outcome
LCS interest
Yes
(n = 24)
Maybe
(n = 36)
No
(n = 21)
LCS clinic visit scheduled
2 (8%)
2 (6%)
0 (0%)
LCS clinic visit completed
1 (4%)
2 (6%)
0 (0%)
Screening or nodule follow-up chest CT completed
3 (13%)
1 (3%)
0 (0%)
Any chest CT completed
5 (21%)
3 (8%)
2 (10%)
Abbreviations: CT, computed tomography; LCS, lung cancer screening.
Discussion
LCS decreases mortality yet is greatly underutilized, revealing a critical need for
strategies to increase LCS among eligible individuals in the United States. Despite
guidelines recommending LCS for high-risk patients, multiple barriers limit its use.
Our pragmatic trial demonstrates that a digital outreach strategy combining an EHR
query with invitations sent via the patient portal has broad reach. Over 85% of patients
in this trial read the portal invitation, and 40% visited mPATH-Lung. These high rates
could partially reflect our inclusion criteria that limited the study to active portal
users, defined as logging in to their accounts within the last 90 days. In addition,
the health system's patient portal automatically notifies patients via email or text
message whenever a new portal message is received. Once the message was opened and
read, the message content indicating a potential for decreased chance of death was
likely a strong motivator to click the embedded link and visit the interactive Web
site. The high response rates we observed should be validated in other studies.
While approximately 50% of Americans use a patient portal, we recognize there are
other more ubiquitous means to reach patients, such as telephone or text message.
Currently, 95% of adults have a cellular telephone with 77% owning a smartphone (including
74% of those aged 50–64), and smartphone ownership is increasing yearly.[35 ] Adding text messages to our digital strategy could increase its reach further. Moreover,
digital outreach strategies require less time and fewer personnel resources than in-person
phone calls.
EHR data, nationally and internationally, is known for relatively poor data quality.[37 ]
[38 ]
[39 ] Constructing EHR selection algorithms require balancing sensitivity (identifying
those who qualify for a service) with specificity (excluding those who do not qualify).
We favored sensitivity over specificity given the minimal burden of having patients
visit mPATH-Lung to determine if they qualify for screening. Therefore, we specifically
designed our EHR algorithm to include ever-smokers instead of more granular smoking
history elements of pack-years or quit-years. This approach led to a large proportion
(72%) of participants determined ineligible for LCS upon visiting the mPATH-Lung interactive
Web site. If this proportion was viewed as too high, our EHR algorithm could be easily
modified to increase its specificity, but more individuals who qualify for screening
would be missed.
Because mPATH-Lung resides on the Internet and can be used at home, it can be scaled
easily without disrupting clinic flow. The Web site currently ends by referring interested
patients to Wake Forest's LCS clinic, but we plan to revise this to assist visitors
from other geographic regions. Our EHR query to identify potential screening candidates
could also be revised to match a health system's data architecture. Our team is currently
working on an approach to automate the sending of the portal messages, a process we
handled manually for this current study.
Many factors influence patients' decisions to pursue LCS. In our study, we found the
risk for developing lung cancer was not significantly different among those who wanted,
did not want, or were not sure they wanted screening, suggesting factors other than
absolute risk alone drive patients' screening decisions. This study was not designed
to evaluate factors influencing patients' decisions to pursue or defer LCS; however,
one survey study and two qualitative studies found that patients consider several
factors in such decisions, including the mortality benefit, chance of false positives,
emotional (anxiety) factors, and practical barriers (costs, transportation).[17 ]
[40 ]
[41 ] LCS decision aids can help patients weigh these factors and make decisions concordant
with their values.[42 ]
[43 ]
[44 ]
The gap between patients who indicated a desire for screening and those who completed
screening in our trial indicates a need for additional patient support. This gap has
been found in other studies of cancer screening decision aids.[27 ] In our study, patients indicated screening preference, but were not able to directly
schedule a screening visit. Allowing patients to schedule their own LCS clinic visits
via the interactive Web site or automatically alerting primary care physicians that
a patient is eligible could facilitate screening, as was seen in another trial in
which patients could “order” their own screening.[29 ]
Our study has limitations. Our use of a single-arm pragmatic design yielded important
information about the feasibility and reach of the digital outreach strategy, but
our ability to examine its effectiveness for screening completion is limited by the
absence of a control group. We are planning a future randomized study to examine this
important outcome. There were also a relatively small number of Medicaid and uninsured
individuals in our study sample. Completion of screening was determined by electronic
review of the EHR and would miss any screening completed outside our institution.
However, given the paucity of screening programs in the region, we believe this is
unlikely. Additionally, the study was conducted within a single health care system
with participants who were active portal users. The effect of the intervention could
be different in different populations. Lastly, the study lacks power to detect differences
within smaller subgroups.
Conclusion
A significant proportion of patients act on health information delivered via the patient
portal. In this pragmatic study, over 85% of patients read the portal messages and
40% acted by visiting the screening interactive Web site. While our study focused
on LCS, this digital outreach approach could be generalized to other health needs.
Importantly, systematically querying the electronic medical record and sending messages
to patients who may benefit from a service requires little staff time. Given the broad
reach and potential low cost of this digital strategy, future research should investigate
best practices for implementing similar systems in diverse health care settings.
Clinical Relevance Statement
Clinical Relevance Statement
Given the high prevalence of smartphone ownership and Internet use, online patient
portals have the potential to broadly engage large numbers of patients in their care.
Our study demonstrates that almost all patients read a health-related message that
was sent to their portal account, and almost half of patients acted on the delivered
information. This digital outreach approach could be applied to any health condition
to inform patients and coordinate care delivery.
Multiple Choice Questions
Multiple Choice Questions
What percentage of U.S. adults with annual household incomes less than $30,000 or
only a high school education have a smartphone?
About 67%.
About 50%.
About 25%.
Less than 10%.
Correct Answer: The correct answer is option a. Smartphone ownership is very prevalent in the United
States. According to a survey conducted by the Pew Research Center in January 2018,
even two-thirds of adults with annual household incomes less than $30,000 or only
a high school education have a smartphone.
What proportion of patients in this study read the patient portal message and visited
the lung cancer screening interactive Web site?
Correct Answer: The correct answer is option b. Almost all patients (86%) read the message that was
sent to their patient portal account, and 40% of patients visited the lung cancer
screening interactive Web site. These findings demonstrate the potential for health
systems to reach large numbers of patients via the patient portal.
