Keywords
tobacco cessation - clinical decision support - physician - primary care - user-centered
design - workflow
Background and Significance
Background and Significance
Tobacco use is a significant cause of morbidity and mortality in the United States.
In 2014, it was estimated that tobacco was associated with 480,000 deaths in the United
States.[1] Costs of tobacco use for the years 2009 to 2012 totaled between $289 and $332.5
billion and direct medical care costs of adults accounted for $132.5 to $175.9 billion
annually.[1] Among the many benefits of tobacco cessation are a decrease in the incidence of
lung and colorectal cancer, cardiovascular disease, and chronic obstructive pulmonary
disease.[2]
[3] Tobacco cessation has been shown to improve overall health care costs and the economics
in households of former tobacco users.[2]
Reports demonstrate that clinician interventions can be effective in assisting patients
with tobacco cessation. The U.S. Department of Health and Human Services strongly
recommends that all clinicians advise patients who use tobacco to quit.[4] A Cochrane review noted that brief tobacco advice from a physician can increase
cessation rates by 1 to 3%.[5]
Among patients, there is a significant interest in tobacco cessation. In 2013, a survey
from the Behavioral Risk Factor Surveillance System found that approximately two-thirds
of smokers quit or try to quit tobacco use annually.[6] Physicians may be missing opportunities to assist patients with tobacco cessation;
prior survey results have shown as many as 80% of current tobacco users do not have
tobacco cessation assistance documented during visits with clinicians.[7]
An opportunity exists to help clinicians promote tobacco cessation using clinical
decision support (CDS). CDS is defined as providing “timely information, usually at
the point of care, to help inform decisions about a patient's care.”[8] In general, CDS tools intended to nudge clinician behavior exist but are challenging
to successfully implement. A 2006 review showed that drug safety alerts were overridden
by clinicians in 49 to 96% of cases.[9] A 2013 study showed that 52.6% of CDS alerts were overridden by physicians in primary
care clinics.[10] In our institution, interruptive alerts shown to physicians or advanced practice
providers (APPs) in 2020 were cancelled 68.6% of the time.
Existing CDS literature regarding tobacco cessation has largely focused on alerts
which automatically generate electronic consults to tobacco cessation services,[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18] print information for patients,[13]
[19] automatically add tobacco use to the patient problem list,[14]
[15] or facilitate ordering of medication therapy.[13]
[14]
[15]
[18]
[20] Some studies, which explored clinical prompting of tobacco cessation counseling,
were intended to be usability-focused or prototypes.[17]
[21] Others have implemented brief counseling training but were unable to integrate into
the provider workflow.[22] One study promoted counseling within the clinician workflow, but did not report
follow-up tobacco cessation rates.[18] This study evaluates tobacco cessation patient outcomes associated with the implementation
of a widely used, user-centered, workflow-integrated, noninterruptive CDS tool designed
to increase clinician awareness of patient tobacco use.
Objectives
Our goal for this quality improvement (QI) project was to increase tobacco cessation
by nudging clinicians using a noninterruptive, note-based CDS tool.
Methods
We implemented a CDS tool in our electronic health record (EHR) (Epic Systems, Verona,
Wisconsin, United States) which adhered to the CDS Five Rights: providing the right
information, to the right person, in the right intervention format, through the right
channel, at the right time in workflow.[23]
[24] During visits with current tobacco users, our intervention dynamically inserted
a selectable list into the Assessment and Plan section of clinicians' notes alerting
clinicians to patients' tobacco use and providing standardized documentation.
Development of this CDS tool was carried out through application of user-centered
design principles. We initially engaged key stakeholders through a series of three
preliminary meetings. Stakeholders included clinicians, medical assistants (MAs),
and quality program managers at Practice A who had aligned interests in improving
quality of care for tobacco use. In interviews, MAs reported regular collection of
patients' tobacco use status. However, clinicians said they were often unaware of
this information as it did not carry forward to their workflow in the EHR. Clinicians
additionally reported that tobacco use status, although desirable information, was
difficult to remember to check due to the sporadic and infrequent number of visits
with active tobacco users. Clinicians also requested an alert for every active tobacco
user at every visit and preferred an intuitive CDS which did not require training
to use.
Using this feedback, we determined the “right information” (patient is a current tobacco
user) for “the right person” (clinician) with “the right information format” (a documentation
aid within the note) in “the right channel” (noninterruptive, note-based) in “the
right time in the workflow” (clinicians' review of the assessment and plan section,
inline). We subsequently met one-on-one with clinicians at Practice A on three different
occasions to gather usability feedback for our iterative design changes. We used a
small-scale demonstration test at Practice A to refine early versions of the CDS tool,
demonstrate usability, and build clinician buy-in before expanding to additional practices.
Clinician training was not provided at subsequently enrolled practices.
The CDS tool was initially launched at Practice A on August 9, 2016, with two subsequent
iterations released on August 25, 2016 and May 1, 2017, based on clinician feedback.
The third and final version of the tool was expanded to three additional practices.
Setting/Population
This intervention was implemented in four metro-Denver Primary Care practices within
a single health care system as part of a QI study. Clinicians eligible to receive
the intervention were attending physicians (Doctor of Medicine or Doctor of Osteopathic
Medicine), APPs (Nurse Practitioner or Physician Assistant), or residents seeing patients.
Patients were included in this intervention if they reported active tobacco use, defined
by the EHR as the use of tobacco products (cigarettes, pipe, cigars, and e-cigarettes)
and/or smokeless tobacco (chew and snuff).
Standard Workflow
As part of the standard rooming workflow at our institution, MAs ask patients if they
are currently using smoking or smokeless tobacco and if so which type(s). If a patient
is a current tobacco user, MAs also ask if they are ready to quit. Lastly, MAs attest
that they have reviewed the tobacco use information with the patient at that visit.
This is repeated at every visit and recorded discretely in the social history section
of the EHR ([Fig. 1]).
Fig. 1 Standard tobacco screening items in the MA workflow. MA, medical assistant.
At the time of this intervention, patients could not self-report tobacco status through
the EHR's patient portal. Clinicians could elect to automatically insert tobacco use
data into their note template.
Intervention Workflow
In the intervention period, MAs continued to screen for tobacco use at every visit.
However, for clinicians who used a template containing the tool, a selectable list
indicating active tobacco use was automatically inserted in the clinician note if
the patient was noted to be a current tobacco user. Current tobacco use was considered
verified if the MA attested to reviewing this information or documented that the patients
reported that they were ready to quit at the current visit. The list was displayed
on its own line in the note, labeled “Tobacco assessment and plan” and contained choices
which facilitated documentation ([Fig. 2]). The patient visit could not be closed without addressing this list in some way,
either by deleting it or choosing a selection from the list. To support this workflow,
ongoing MA training was provided to reinforce the importance of addressing tobacco
use history at every visit.
Fig. 2 Selectable list for current tobacco users, expanded to show documentation selection
options.
Study Design and Measures
To evaluate the impact of the tobacco cessation intervention on tobacco cessation
rates, we conducted a retrospective analysis of patient-level data obtained at four
practice sites from January 1, 2017 to December 31, 2019. We utilized a pre–post study
design where data were collected from all sites both under the standard workflow (control)
and after the intervention workflow was available. The implementation of the intervention
was staggered across the sites, occurring at the first site in August 2016 and at
the last site in December 2018.
Patients were included in the analysis if the EHR identified them as current tobacco
users and at least one of the following two conditions was met: (1) the MA attested
to reviewing the patient's tobacco status and (2) the MA documented that the patient
was ready to quit at the visit. The primary encounter date was the first eligible
visit for a person based on the eligibility criteria. The primary outcome was time
to first reporting of tobacco cessation. Patients were asked if they currently use
tobacco at every visit as part of the standard MA workflow. Tobacco cessation was
defined as the first follow-up visit at which the patient tobacco status changed from
current user to nonuser when reviewing screening questions with the MA. If an individual
was not seen again in the study period or if their tobacco status was not reviewed
as defined above, it was assumed their tobacco status had not changed. The primary
predictor of interest was the intervention group (intervention/control). Differences
in cessation rates between the two groups were examined at 90, 180, and 365 days.
Potential Confounders
Multivariable analysis was adjusted for patient age, self-reported race, identified
gender, location, and the site at which the patient had the initial visit. Patient
location was identified as county based on zip code, which was then further classified
using Rural-Urban Commuting Area (RUCA) codes as urban (RUCA 1–3) and rural (RUCA
4–10).[25]
Outcomes
Time to event for individuals who experienced cessation prior to the end of the study
was the time from baseline (first visit in the study) to the date of cessation. Time
to event for patients who did not return for a follow-up visit or did not experience
cessation during the study period was defined as the end of the study period or the
last follow-up date, whichever occurred first. Patients who had an eligible visit
in both the control and intervention periods had two entries in the data set and their
last follow-up date in the control period was the date of the first eligible visit
in the intervention period.
Analysis Methods
Kaplan–Meier curves were used to compute tobacco cessation probabilities for key time
points of interest at 90, 180, and 365 days. A Cox proportional-hazards model was
employed to assess the association between intervention and tobacco cessation, adjusted
for patient age, gender, location (rural, urban), and practice site. Clustering due
to repeated measurements on patients who were seen in both the intervention and control
conditions was accounted for using robust standard errors. This time-to-event analysis
approach allowed for the conservative assumption that individuals who do not return
for a follow-up were still considered tobacco users.
Results
There were 5,644 patients included in the analyses. The median patient age was 45
years (interquartile range: 33–57) and 60% of patients were males. The majority of
patients (>99%) lived in an urban location and were white (63%) ([Table 1]). Over a median follow-up time of 16 months, 1,161 patients reported a tobacco cessation
event.
Table 1
Descriptive characteristics of cohort (n = 5,644)
|
Characteristics
|
N (%)
|
|
Site
|
|
|
Practice A
|
1,318 (23%)
|
|
Practice B
|
1,988 (35%)
|
|
Practice C
|
1,466 (26%)
|
|
Practice D
|
872 (15%)
|
|
Age, y, median (IQR)
|
45 (33–57)
|
|
Male
|
3,359 (60%)
|
|
Patient location
|
|
|
Urban
|
5,613 (99.5%)
|
|
Rural
|
31 (0.5%)
|
|
Race
|
|
|
White or Caucasian
|
3,543 (63%)
|
|
Black or African American
|
947 (17%)
|
|
Asian
|
159 (3%)
|
|
American Indian and Alaska Native/Native Hawaiian
|
30 (0.5%)
|
|
Other/more than one race
|
705 (12%)
|
|
Patient refused/unknown
|
260 (5%)
|
From the Kaplan–Meier analysis ([Fig. 3]), there is a difference in the survival curves for tobacco use between the two groups
(p = 0.002). At 90 days, the probability of tobacco use was 95.2% in the intervention
group and 96.1% in the control group, an absolute difference of −0.9%. At 180 and
365 days, the absolute difference in the probability of tobacco use between the intervention
and control groups was −1.0 and −2.8%, respectively. In the multivariable Cox proportional-hazards
model adjusted for patient age, race, gender, location, and practice site, the intervention
group had significantly greater risk of tobacco cessation compared to those in the
control group (hazard ratio: 1.22, 95% confidence interval: 1.08–1.36; p = 0.001) ([Supplementary Table S1], available in the online version). There was no significant evidence that gender
modified the effect of the intervention on tobacco cessation (p = 0.71). Of the 254,742 encounters that were recorded at the four sites during the
study period, there was a significant difference in the rate at which MAs reviewed
a patient's current tobacco use or readiness to quit in the control and intervention
arms (84 vs. 94%; p < 0.001) ([Supplementary Table S2], available in the online version).
Fig. 3 Tobacco use curves by treatment arm for 90, 180, and 360 days following the index
visit.
Discussion
Using a minimalistic, dynamic CDS tool, we observed a significant improvement in tobacco
cessation rates among patients at intervention clinics. Our intervention, which inserted
timely, verified tobacco use information into the progress notes of patients with
active tobacco use, provided an “alert” within the clinicians' workflow. While the
alert had to be addressed, it was noninterruptive and desired by clinicians, which
we determined through our user-centered design process. This tool utilized “right
information” (patient is a current tobacco user) for “the right person” (clinician)
with “the right information format” (a documentation aid within the note) in “the
right channel” (noninterruptive, note-based) in “the right time in the workflow” (inline,
clinicians' review of the assessment and plan section).
We suspect that the observed tobacco cessation rates are due to increased clinician
attention to patients' tobacco use through successful application of the CDS Five
Rights. Clinicians who are aware of patient tobacco use “in the moment” can then draw
on their training to deliver brief advice, refer to tobacco cessation counseling,
improve documentation, or write appropriate prescriptions. While this report does
not capture how clinicians specifically chose to apply the tobacco use data, we demonstrate
the importance of successfully delivering this desired and clinically relevant information.
These findings are significant because of the potential magnitude of clinical effect
if applied to a larger population. The absolute difference in cessation rates from
a baseline of −2.8% tracks closely with the expected cessation rate of 1 to 3% due
to brief tobacco cessation advice from a clinician alone.[5] In 2021, our system had over 1,430,000 active patients of whom over 140,000 reported
active tobacco use. If our intervention were deployed across our system, there would
be nearly 4,000 fewer tobacco users in 1 year. It is known that changing quit rates
by even small percentages will result in societal benefits including reduction of
health care utilization, morbidity, and mortality.[2]
[3] Giving clinicians tools which measurably influence tobacco cessation rates could
have a large overall impact.
Anecdotally, we did not hear concerns from clinicians about extra work or burden of
using the tool. Users reported that the workflow would be enhanced if the tool were
in the Subjective section of the note instead of the Assessment and Plan section to
improve chances the tool would be seen during the face-to-face portion of the visit.
They also felt the detailed documentation the tool provided was unnecessary and simpler
text would have a similar impact. This feedback will be implemented into future versions
of the tool.
This study has many strengths. The CDS tool used reliable data and only prompted clinicians
when tobacco status had been verified at the current visit. We intentionally created
a CDS tool which did not require clinician training to use, decreasing the burden
of implementation. The long study period allowed us to have a relatively large data
set and successfully track patients over time. We were able to employ the intervention
across clinicians of all training levels in both General Internal Medicine and Family
Medicine, testing the applicability of the tool across multiple clinician practice
types. MA workflow remained the same across both the control and intervention periods,
limiting potential bias. Our model also promotes the U.S. Preventive Services Task
Force recommendation regarding clinician-driven tobacco cessation by identifying current
tobacco users for clinicians during the visit to facilitate face-to-face advise and
interventions.[26]
This study also includes some limitations. As the tobacco use status of nonreturning
patients was not assessed, our data may underestimate the impact of the intervention.
The intervention was limited by reliance on MA documentation of tobacco status and
inclusion of the CDS tool in clinicians' notes. MA documentation of tobacco status
did significantly increase during the intervention period, though it is unclear what
impact this increase had on the cessation outcome. This study did not control or measure
clinicians' specific tobacco treatment interventions which may have provided more
insight as to how the tool was used. Additionally, this QI study was conducted at
only four sites in a Colorado health system whose population was predominantly urban.
Patient characteristics such as ethnicity, insurance category, and history of chronic
illness are potential confounders that were not available in this data set and thus
not adjusted for in multivariable analyses. While we demonstrate a positive association
between the intervention and tobacco cessation in this quasi-experimental study, randomized
controlled trials are required to establish causation.
The implementation of this CDS tool was designed as a QI activity to improve patient
care. As such, data collection and analysis were limited to the minimum needed to
evaluate the outcome of tobacco use. Future work might study this tool in a larger
number of clinics to assess generalizability, analyze differences in clinician use
of tobacco cessation data, evaluate relationships between tobacco cessation and patient
comorbidities, explore history of prior quit attempts by patients, and examine associations
between clinician documentation and quit rates.
Conclusion
Tobacco use is a significant cause of morbidity and mortality which can be reduced
by even brief clinician interventions. By raising awareness of tobacco use in clinicians'
notes using principles of the CDS Five Rights and user-centered design, cessation
rates were improved. This intervention method deserves further research to continue
integration of end-user feedback and expansion to a larger population of clinicians
and patients.
Clinical Relevance Statement
Clinical Relevance Statement
This study demonstrates a successful implementation of CDS, which adheres to the CDS
Five Rights. Employing the user-centered design and following this template, readers
should be able to implement similar targeted CDS tools within their health systems.
Multiple Choice Questions
Multiple Choice Questions
-
Which of the following best describes the Five Rights of effective clinical decision
support?
-
Right information, using the right wording, on the right patient, through the right channel, at the right time
-
Right patient, to the right clinician, in the right format, through the right channel, at the right time
-
Right information, to the right person, in the right format, through the right channel, at the right time
-
Right image, at the right size, shown to the right person, using the right colors, at the right time
Correct Answer: The correct answer is option c. The best practice framework for clinical decision
support tools includes these Five Rights: the right information, to the right person, in the right intervention format, through the right channel, at the right time in the workflow for decision or action.
-
Which of the following is a principle of user-centered design?
-
Design and test usability iteratively
-
Slowly gather user feedback
-
Inform users of why the product is ideal
-
None of the above
Correct Answer: The correct answer is option a. The principles of user-centered design involve including
users throughout the CDS design process, gathering rapid cycle feedback, and making
iterative changes based on user testing.
-
When applying user-centered design and the CDS Five Rights to a clinician alert for
current patient tobacco use …
-
Tobacco cessation rates decrease
-
Tobacco cessation rates increase
-
Tobacco cessation rates remain the same
-
Tobacco cessation rates vary
Correct Answer: The correct answer is option b. When compared to the control group, we saw increased
tobacco cessation rates in intervention practices following the implementation of
an alert in clinicians' notes that included information on patients' active tobacco
use status.
-
When applied to a larger population of patients and clinicians, this intervention
could potentially impact patients'…
-
Health care utilization
-
Morbidity
-
Mortality
-
All of the above
Correct Answer: The correct answer is option d.
