Keywords
clinical decision support - heart failure - computer systems development - primary
health care - adrenergic β-antagonists
Background and Significance
Background and Significance
Heart failure (HF) is an increasingly prevalent problem.[1] One of the treatments that has shown to improve clinical outcomes in HF patients
with depressed ejection fraction is the use of β-blocker at optimal doses; their use
is currently a class 1A recommendation per the American College of Cardiology and
American Heart Association guidelines.[2]
[3] The guidelines also recommend that “physicians should make every effort to achieve
the target doses of the beta-blockers shown to be effective in major clinical trials.”[4]
[5]
[6]
[7] Patients with HF and depressed ejection fraction have better symptom management
and survival when treated in accordance with the guidelines.[8] Beta-blocker use at optimal doses has also been associated with a reduction in hospital
readmissions.[9]
[10]
[11]
[12]
[13]
Unfortunately, a large portion of patients with HF and depressed ejection fraction
are not receiving guideline-recommended medication treatment[14]—incorrect β-blocker drugs are used, or the dosage is too low, or there is no use
of β-blockers.[7]
[15]
[16] A recent global survey found that although 87% of HF patients with reduced ejection
fraction were on β-blockers, only 15% were at the target dose.[17]
The decision for determining the best target dose of β-blocker is complex. However,
prior work has explored the use of clinical decision support systems (CDSS) in supporting
management of HF by primary/general practitioners.[18]
[19]
[20] Leslie et al[18] identified challenges to HF treatment including inadequate support for uptitration
of β-blockers, and how nonmedical needs of patients can complicate treatment decisions.
Toth-Pal et al[19] found that CDSS can impact medication decisions, but also that the time involved
in using a CDSS is a concern for providers. Smeets et al[21] report that the treatment of HF by primary care providers (PCPs) is hampered by
a lack of access to specialized knowledge; one suggestion is to use CDSS to integrate
evidence-based guidelines into clinical practice.
The purpose of this article is to report on the requirements development and design
of a CDSS tool that will inform and motivate primary care clinicians in making the
decision about β-blocker use for patients at risk for undertreatment. We focus on
primary care because of the greater number of and access to PCPs compared with cardiologists,
and the greater likelihood that a patient who is not seeing a cardiologist may be
receiving suboptimal treatment. The tool we present here is based on a software developed
to read ejection fraction from free text reports in electronic health records (EHRs)[22] to identify HF patients with depressed ejection fraction who might qualify for high
doses of β-blockers but are not receiving them.[23] It is different than prior CDSS tools focused on HF[18]
[19] in that (1) it is focused specifically on β-blocker titration for reduced ejection
fraction patients; and (2) it identifies if a given patient is at risk and pushes
a notification to the provider at an appropriate time, rather than requiring the provider
to recognize a problem and seek out the tool.
Simply giving PCPs information on potentially undertreated HF patients may not lead
to better use of β-blockers.[24] There are challenges related to both the task of making medication decisions for
HF patients, and to the context in which the decision is made. In HF patients with
depressed ejection fraction, titration to optimal doses, and maintenance at those
doses, can be a complex process. Fluid retention needs to be carefully monitored.
Beta-blocker use can affect blood pressure and heart rate, and requires careful management
in patients being treated for those conditions as well.[25] These demands occur in the context of time constraints on primary care with many
clinical reminders for multiple conditions, making it challenging to spend time managing
chronic diseases.[26]
[27] Additionally, decisions about HF management can be influenced by uncertainty about
clinical practice, personal experience with HF management, and the availability of
resources.[28]
These factors are important in adoption and observance of guidelines.[29] Understanding these factors, and considering them in the design of the CDSS, is
critical for developing a tool that both meets the needs of the end user, and is designed
with an understanding of the context in which it will be used. Fortunately, there
are existing frameworks and design research methods that can help with the development
aspects.
The Promoting Action on Research Implementation in Health Services (PARIHS)[30] implementation framework emphasizes characteristics of the clinical contents and
the organizational context in implementation. It includes three important areas—evidence,
context, and facilitation—all of which should be used to translate evidence-based
guidelines into actual clinical practice.
Part of the organizational context and facilitation of clinical work depends on the
health information technology (HIT) and its integration with the organization. This
is included in the HIT sociotechnical framework,[31] based on the Systems Engineering Initiative for Patient Safety model.[32]
[33] This framework lists several areas that are important to consider regarding fit
of HIT tools in the sociotechnical context of their operation, including clinical
content, human–computer interface, and workflow and communication.
The Cabana framework[34] looks at provider-related dimensions to explain barriers related to adherence to
clinical practice guidelines. It looks at gaps in the areas of knowledge, attitude,
and practice required for following the practices recommended by the guideline.
These frameworks have been used for evaluation and implementation of solutions, and
the identification of barriers and facilitators;[35] however, their use as part of requirements development is much less evident. One
exception is Sheehan et al's[36] use of sociotechnical analysis to inform the design of a CDSS for pediatric head
trauma.
The three frameworks fit together to help identify how the provider must execute the
practice recommended by the guideline in the sociotechnical context of the organization.
Cabana focuses on providers' adherence to guidelines, especially in terms of knowledge
and attitude.[34] PARIHS focuses on the content and context of the work involved in following the
new practice.[30] The HIT sociotechnical model focuses on how the overall HIT-enabled health care
system supports safe and effective care delivery.[31] Collectively, these three frameworks provide an approach that incorporates the individual
provider, the work context, and the overall system.
To help translate these demands into functions performed by the provider, and therefore
design requirements for the CDSS, we used the cognitive tasks analysis (CTA) methodology.[37] This is a family of methods for studying the patterns of cognitive work in real-world,
complex contexts, to identify challenges for practitioners and opportunities to provide
better support. Such analysis is important for generating design requirements that
will address the challenges of the practitioners—in this case, the challenges faced
by providers in implementing the β-blocker guidelines related to β-blocker use and
performing dose titration when appropriate for qualifying HF patients. Developing
systems by focusing on meeting the cognitive needs of the users is recommended to
facilitate adoption.[38]
[39] This is especially important for CDSS.[40]
CTA includes study of both the cognitive work of the practitioner and the essential
characteristics of the work domain.[41] The use of the frameworks to guide and enhance the CTA is a way to incorporate into
the CTA some important theoretical insights about processes and challenges in the
domain of health care. The frameworks provide an understanding of general implementation
and adoption issues in the complex sociotechnical context of HIT-enabled health care
systems. The CTA-based requirements elicitation process incorporates these issues
along with the specific cognitive challenges of HF treatment and β-blocker titration,
pointing out opportunities (requirements) for supporting the cognitive work of the
providers (See [Fig. 1]).
Fig. 1 Cognitive task analysis (CTA)-based requirements elicitation process informed by
theoretic frameworks.
Objective
The purpose of this study is to develop (1) requirements and (2) a prototype design
for a clinical decision support tool. This tool gives PCPs targeted information on
HF patients with depressed ejection fraction who quality for guideline-recommended
β-blocker treatment but are not receiving it. The requirements are generated via qualitative
data collection and analysis based on CTA methods[42] enhanced through the integration of the PARIHS, HIT sociotechnical, and Cabana frameworks.[43] The goal is to address issues of adoption and implementation as early as possible
in the design process, instead of trying to address them only after a tool has been
developed. A secondary objective of the study is to conduct an initial assessment
of the design via a pilot usability evaluation.
Cognitive Task Analysis
We conducted CTA interviews with providers, and analyzed the results, to inform the
design requirements (described in this section). Afterwards, these results were used
to help develop a prototype Clinical Reminder (the “Clinical Reminder Prototype” section).
This was reviewed by a few key informants, and refined based on their feedback. The
revised prototype was then usability tested with a few PCPs (the “Usability Review”
section).
Cognitive Task Analysis Methods
Setting and Participants
The setting was the Veterans Health Administration (VHA). The interviews were conducted
at three outpatient clinics from the same regional network. One was in a large city,
one in a medium-sized city, and one in a rural area. All the clinics had implemented
the VHA's medical home model for primary care, called Patient Aligned Care Team (PACT).[44] Our analysis of the larger two of these sites found that only 62% of HF patients
were receiving a β-blocker, which is similar to the findings of an earlier nationwide
Veterans Affairs (VA) study.[45]
We recruited three types of PACT members: PCPs, clinical Pharmacists (PharmDs), and
Registered Nurses (RNs). By interviewing PharmDs and RNs, who play important roles
in collecting information about and providing treatment for HF patients, we were able
to more fully understand the context in which PCPs are doing clinical evaluations
and assessing medication options. We used convenience sampling based upon participant
availability during the days of our visits to each facility. Institutional review
board approval was obtained for this study.
Data Collection
We developed specific interview guides for each provide role (PCP, PharmD, RN). The
interview guides were pilot tested with three providers (one PCP, one PharmD, and
one RN) at another VHA site. The guides included questions on the provider's role
in HF management, and how the provider learned of information about specific patients'
care needs. The guides for the PCP and PharmD included questions about their knowledge
of the guideline, attitudes toward such guidelines, experience with systolic HF management,
and familiarity with β-blocker titration. Questions about organizational environment
and EHR issues were also included. Thus, the guides incorporated elements from the
Cabana, PARIHS, and sociotechnical frameworks.
To ensure responses were grounded in actual experience of work-as-performed, we also
asked participants to walk us through their management of a recent HF case (as per
the critical decision method[46]), with emphasis on what decisions they needed to make and what factors they needed
to consider. The key relevant interview questions are presented in [Table 1].
Table 1
Representative key interview questions
|
Role
|
Question
|
|
RN
|
How about for management of patients with chronic conditions, like CHF? What is your
role?
|
|
PharmD
|
In terms of managing medications for outpatients with congestive heart failure, what
is your role?
• What is your scope of practice relative to that area? Which orders, if any, need
to be cosigned by the PCP?
• In terms of the management of a patient with CHF, how do you interact with the PACT
teamlet?
|
|
PCP
|
Have you ever had a patient with systolic CHF who was hospitalized and later discharged
for exacerbation of systolic CHF? Yes? So, do you recall…
• No? Well, think about any cases you may have had regarding a patient with a chronic
conditions (such as CHF) being discharged.
• How did you/would you learn about this patient being discharged?
• Let's say you do have a patient who was recently discharged after being hospitalized
for exacerbation of systolic CHF… What was/would be the process for follow-up?
• What challenges are there that make it harder to accomplish this postdischarge follow-up?
|
|
PCP, PharmD
|
In terms of the management of the patient's systolic CHF, what were you/would you
be looking at or thinking about?
• What medications would you look at?
o Are β-blockers a medication type that you have used with systolic CHF patients?
o Which β-blockers have you used for systolic CHF patients?
• How to you determine the dosage for the patient?
o Do you keep them on one dose, or do you titrate the dose at each visit? Why? When
do you titrate (only if BP is elevated, or as long as BP is not too low)?
• What challenges make it difficult for you to achieve use of β-blockers with your
patients?
|
|
PCP
|
In general, do you find guidelines helpful? Do you have any concerns about the role
of guidelines in primary care?
|
|
PCP, PharmD
|
I'd like to ask you about guidelines for CHF management. Are you familiar with CHF
management guidelines? Has someone told you about them?
• Can I ask you a couple of things about the guidelines? It is just to see what info
we might want to repeat in any reminder notification, or what would be redundant.
o For the β-blockers you have used for systolic CHF (name of β-blocker), do you know the target doses recommended in the guideline?
o Do you know the purpose of using β-blockers in systolic CHF patients?
|
|
PCP, PharmD
|
Do you have any concerns about the CHF guidelines and how they relate to your systolic
HF patients?
• Do you have any concerns about prescribing β-blockers to systolic CHF patients?
|
|
PharmD
|
Can you walk me through the titration process? Who would be involved? How would information
be coordinated?
|
|
PCP
|
Is titrating β-blockers to the guideline recommended doses something you feel comfortable
with doing for a systolic CHF patient, or is that something you would have a pharmacist
or cardiologist or other person do?
• Can you tell me what is it about [pharmacists, cardiologist] that makes them a better
choice for handling that?
|
|
PCP, PharmD
|
Are there any factors or issues which prevent you from titrating to the guideline
recommended doses?
• Are there tools you use to help with titration?
|
|
RN
|
We are considering situations where a patient with a chronic condition gets discharged
from the hospital, but treatment for that chronic condition is not sufficiently reviewed
and updated.
• What factors would contribute to those sorts of problems in an outpatient clinic?
Bigger panel, less staff, more complex patients…
• Think about a time or place where the working conditions were more difficult. Under
those conditions, how did you monitor and keep track of things?
• Under those conditions, how might you have responded to that type of notification—about
a discharged patient not getting guideline recommended care for his chronic condition?
|
|
PCP
|
What determines if a patient gets CHF treatment primarily from a CHF clinic, a cardiology
specialist, or by you and your PACT?
|
|
PCP, PharmD, RN
|
Imagine you (or someone in your PACT) gets a notification that says that our software
has detected that one of your patients has recently been discharged from the hospital
(admitted for CHF exacerbation)—and that it looks like they may qualify for being
titrated up to guideline recommended levels of β-blocker.
• Tell me what that message means to you.
• What would be done? What might be different as compared with the process you described
before?
• What would you look at to evaluate the patient for treatment with β-blockers?
• What information would you want from the notification?
|
|
PCP, PharmD
|
You mentioned some of the factors that make it harder to manage postdischarge follow-up,
and to get patients on β-blockers. When those factors are in play—when there is more
workload, when communication and engagement with the patient is harder, what do you
do different in terms of postdischarge follow-up for CHF? In terms of medication management?
How and when do you look at: discharged patients? Patients being titrated? Patients
with CHF? How are other PACT members involved?
• Under those conditions, how would you respond to that type of notification—the one
about a CHF patient who wasn't getting guideline recommended β-blocker treatment?
|
Abbreviations: CHF, congestive heart failure; PACT, Patient Aligned Care Team; PCP,
primary care provider; RN, Registered Nurse.
The interviews were semistructured, and conducted in person by two team members (M.S.
and C.B.) who took notes. They took place onsite in vacant rooms or the participant's
private office. They were audio-recorded and transcribed. Each interview lasted between
30 and 60 minutes.
The interviews were designed to identify the factors related to the decision about
whether or not to titrate a patient to target doses of β-blockers. Note that during
the interview phase of the project, the decision to develop a clinical reminder had
not been reached.
Data Analysis
We analyzed the data using the framework approach.[47]
[48] Two coders (M.S. and C.B.) independently reviewed the data. During the initial informal
pass, the coders followed an inductive approach based on the literature on health
care teamwork[49] and PACTs,[50]
[51] and also looked for emergent themes. After discussions among the interdisciplinary
team, the coding approach was refined, following a deductive approach based on the
Cabana, PARIHS, and HIT sociotechnical frameworks. With this approach, the PARIHS
and HIT sociotechnical frameworks helped to identify organizational factors (reflecting
the context of work), as well as patient and clinical factors (reflecting the content
of work). The Cabana framework helped to identify the factors associated with the
providers' knowledge, skills, and attitudes.
The two coders used Atlas.ti to code the data. A subset of transcripts was coded by
both coders independently, to establish intercoder reliability. Afterwards, the remaining
transcripts were coded only by one or the other coder. In an iterative process, the
coding results were jointly reviewed and interpreted. As an additional cross-check,
each coder evaluated the other coder's results by comparing them to their own interview
notes and postinterview assessments (which served as an independent summary for comparison).
The results were reviewed by our interdisciplinary research team (which included specialists
in cardiology, psychology, cognitive systems engineering, and medical informatics).
The team identified patterns in the results regarding: cognitive work of β-blocker
titration decisions; knowledge and attitudes; work context; and system barriers and
facilitators. These patterns helped frame design requirements.
Cognitive Task Analysis Interview Results
For the CTA interviews there were a total of 17 participants: 7 PCPs (2 physicians,
5 nurse practitioners), 5 PharmDs, and 5 RNs. Note that this is a typical sample size
for naturalistic CTA studies.[52] The sample of PCPs has only two physicians, but the ratio of physicians to nurse
practitioners is reflective of the staffing in primary care at these sites. Overall
average number of years in practice was 18 (standard deviation, 9.6); PCPs had 17,
PharmDs had 6, and RNs had 26 years on average. [Table 2] summarizes the main issues mentioned in the CTA interviews.
Organizational and Patient Factors
Participants mentioned various organizational factors that affect titration and other
ways of managing systolic HF. Most PCPs did not refer their patients to pharmacists.
One pharmacist indicated that most PCPs see pharmacists as useful for chronic disease
management, but that β-blocker titration was not a common reason for referral. Some
PCPs were of the opinion that most HF patients were working with cardiologists, meaning
that the PCP would leave cardiac issues alone. Also mentioned was how patients who
use both private specialists along with VA care presented challenges regarding accessing
and sharing medical record information.
There was uncertainty and inconsistency about which of the three approved drugs (metoprolol
succinate, carvedilol, and bisoprolol) was available in the providers' VA sites, and
which were available only as a second-line treatment or not at all.
Resource limits and time pressures were often mentioned. An RN noted how PACT in theory
should utilize RNs for chronic care management, but in practice the RNs were too busy
with other responsibilities and did not have time for preventative and chronic care
management. One PCP mentioned that the time burden of seeing a patient every 2 weeks
for titration was a difficulty. The burden for patients, especially those living far
away from the clinic, or with transportation challenges, was also raised.
Other patient factors brought up included the impact of β-blockers on energy levels
and functional capacities, as well as the potential instability of patients (in terms
of volume status and other aspects of compensatory capacity) and how this was affected
by dietary issues. These make β-blocker titration more complicated.
Guideline and Titration Attitudes and Knowledge
Some PCPs expressed some hesitancy at β-blocker titration. One had developed a risk-averse
strategy of referring to specialists when possible. Another had a specific dose limit
beyond which cardiology would be invited to take over. Similarly, some PCPs reported
that they would refer to pharmacists or cardiologists for complex patients.
All of the PCPs knew of the guideline, but many showed an incomplete understanding
of the rationale for the guideline, as well as the approved drugs and target doses.
Most had used one or two of the approved medications. Some were unclear on which version
of metoprolol was approved, thinking it was tartrate instead of succinate.
Most expressed a positive opinion of clinical guidelines overall in helping to provide
quality care. A few also shared concerns that emphasis on this clinical guideline
could contribute to aggressive titration without due consideration of side effects.
Clinical Data
Collectively, a large number of clinical factors were identified as important to consider
when assessing clinical need for and potential tolerance of high β-blocker dosing.
These included:
-
Age.
-
Ejection fraction.
-
Other cardiac data (blood pressure, heart rate).
-
Co-morbidities (diabetes, renal problems, chronic obstructive pulmonary disease [COPD],
asthma).
-
HF symptoms, functional status.
-
Patient stability (indicated by trends in weight, blood pressure, heart rate).
This last point reiterates how knowing the date of the diagnosis or measurement was
important to how it should be interpreted. It was also noted that accessing ejection
fraction, and getting data from various time points to assess stability, are cumbersome
tasks to do in the EHR ([Table 2]).
Table 2
Main issues from CTA interview results
|
Organizational factors (reflecting PARIHS and HIT sociotechnical context of work)
|
• Underutilization of pharmacists
• Concern about interfering with plans of cardiologist or private specialist
• Uncertainty and inconsistency about approved β-blockers on formulary
• Shortage of time and staff for chronic care management and titration
|
|
Patient factors and clinical data (reflecting PARIHS and HIT sociotechnical content
of work)
|
• Impact of β-blockers on patient functional capacity
• Lack of patient stability
• Important clinical data to consider in decision (includes age, comorbidities, cardiac
functioning)
|
|
Provider knowledge, skills, attitudes (reflecting Cabana framework)
|
• Risk or difficulty threshold of when to refer to specialist
• Partial understanding of rationale for guideline
• Positive opinion about clinical guidelines in general
|
Abbreviations: CTA, cognitive task analysis; HIT, health information technology; PARIHS,
Promoting Action on Research Implementation in Health Services.
Clinical Reminder Prototype
Clinical Reminder Prototype
Design Process
Working as a team, we used the analysis of the interviews and the patterns we found
to identify the needs of the providers during consideration of β-blocker titration
for systolic HF patients. These information, workflow, and coordination needs were
used to generate requirements for the functioning of the notification and decision
support tool. Based on the requirements, we determined that the tools would best be
implemented as a VISTA clinical reminder. These are large dialogue windows that appear
when a provider is starting a patient visit note (see [Fig. 2]: example of typical clinical reminder), usually to remind providers to complete
certain screening tasks that are due for that patient.[53]
Fig. 2 Example of typical clinical reminder.
The team generated various design concepts for clinical reminders to meet the requirements,
eventually narrowing it down to a set of concepts that were used to create an initial
nonfunctional prototype. This prototype was reviewed by team members and refined accordingly.
Design Results
Key informants from the three sites gave us feedback on the initial design (described
in the next section). They identified some issues with the wording at the top, and
the formatting of dates. Based on this feedback we made a revised prototype, show
in [Fig. 3]. Our design incorporates several features to meet the needs identified through the
framework-enhanced CTA interview. These are identified by the numbered callouts in
[Fig. 3].
Fig. 3 Prototype used in usability pilot test.
-
The evidence for the guideline is briefly presented at the top.
-
The latest ejection fraction results are provided, including the date and source of
the measurement.
-
The patient's current β-blocker medication and dose, if any, are provided.
-
The three approved medications and their target doses are listed.
-
The provider can make a quick decision if their patient does not qualify. In this
case, only minimal interaction is needed to satisfy the clinical reminder. This means
that if providers do not need to consider β-blocker titration, the time imposition
is minimized. Further interaction is only for cases where assessment for titration
is warranted.
-
Data from the last few measures of weight, blood pressure, and heart rate are presented
to help the provider assess stability.
-
Links to resources on how to titrate are provided.
-
The provider can make an order for β-blockers from the clinical reminder.
-
The provider can also request a clinical pharmacy referral.
Usability Review
Usability Review Methods
We showed the prototype to key informants at the three sites, and asked them to walk
through the reminder as if they were seeing it for a patient. Their feedback was used
to further refine the prototype. The revised version was subsequently evaluated via
a small pilot usability test, involving a scenario-based walk-through assessment done
remotely from the participating sites. The PCPs were given mock-ups of the prototype
showing data and providing relevant clinical notes for real patients (anonymized records
that were used with proper approval). These mock-ups were Portable Document Format
files with internal linking but no other functionality. There were two scenarios,
both involving older patients with suppressed ejection fraction HF and comorbidities
including COPD, who were recently discharged after being in the hospital due to HF
exacerbation. One was a true positive in which the patient qualified for guideline-recommended
levels of β-blocker but was not receiving it. The other scenario was a false positive
in which the patient did not qualify because of blood pressure issues. Information
was elicited via think-aloud and probe questions.
Further evaluations have been done (see Refs. [54 and 55]). However, the clinical reminder has not yet been implemented.
Results of Usability Review
Three key informants (1 PharmD, 2 MD PCPs) participated in the informal usability
review of the initial prototype. They expressed positive opinions about the general
design approach. They also provided feedback about specific functional and layout
issues. These recommendations were implemented in the revised prototype ([Fig. 2]).
Three PCPs (1 NP, 2 MDs) participated in the scenario-based pilot usability test.
These participants were able to make sense of the clinical reminder and use it to
make decisions about β-blocker titration for systolic HF patients with depressed ejection
fraction. The PCPs noted the utility of providing the list of approved drugs and target
doses, and the recent history of weight, blood pressure, and heart rate.
Discussion
In this article, we describe a process for using implementation and sociotechnical
frameworks as part of a CTA-based requirements development process for a clinical
reminder concerning underuse of appropriate β-blocker medications for HF patients
with depressed ejection fraction. Our design includes several functions that help
support the needs of the providers and facilitates adoption. Unlike prior HF CDSSs,[18]
[19] this one is a clinical reminder that appears without any action by the provider.
However, it only appears when the text extraction software has identified the patient
as lacking guideline-recommended β-blocker treatment despite having reduced ejection
fraction. This design is congruent with recommendations to support both automatic
well-learned cognitive processing while at the same time providing tools to attend
to the issue if needed. Combining both automatic and deliberative processing has been
recommended as an approach to CDSS design.[40] The clinical reminder presents the evidence-based rational for the treatment with
β-blocker which activates practice values. When providers are familiarized with the
evidence, it can lead to better attentiveness to the clinical reminder.[56] It facilitates the execution of rapid, well-learned habits among providers who have
sufficient experience in deciding upon and initiating β-blocker titration, thereby
supporting automatic, pattern-matching cognitive processing. At the same time, it
supports providers who still require slower, more explicit processing to make the
decision and start the titration process.[40]
[56] In this way, it supports the integration of these different cognitive processes.
Additionally, there is an open text field for providers to describe reasons why β-blocker
titration may not be applicable to their patient. This avoids constraining the provider's
options, and it captures data that can be used by system administrators to refine
the algorithm to reduce false positives. Considering these design features, and the
results of the initial evaluation, we believe our clinical reminder prototype addresses
the usability and integration problems that have caused problems with other clinical
reminders.[57]
Many of the strengths of the design are due to our innovative requirements elicitation
approach, using CTA enhanced with the PARIHS, HIT sociotechnical, and Cabana frameworks.
Had we followed the same clinical reminder format used for simpler decisions, the
design would have only had a statement that the patient was not getting the right
β-blocker type and/or dose, and checkbox options to either initiate titration or indicate
that the patient does not qualify for β-blocker titration.
Had we done a CTA but without integration of the PARIHS, HIT sociotechnical, and Cabana
frameworks, the design would likely have still included some clinical data and treatment
guideline information, but most likely not many other design features. These are presented
in [Table 3].
Table 3
Design features resulting from application of frameworks
|
Design feature
|
Frameworks
|
|
The rationale for guideline
|
PARIHS
Cabana (attitude)
|
|
The organization to allow quick resolution for false positives, and for providers
who are very familiar with deciding upon and initiating β-blocker titration
|
HIT sociotechnical (time pressures)
Cabana (practice)
PARIHS (facilitation)
|
|
The upfront presentation of β-blocker doses
|
Cabana (knowledge)
|
|
The links to resources on how to titrate
|
Cabana (knowledge)
PARIHS (facilitation)
|
|
The option for consult to pharmacist
|
HIT sociotechnical (workflow)
PARIHS (facilitation)
|
Abbreviations: HIT, health information technology; PARIHS, Promoting Action on Research
Implementation in Health Services.
The decision to present the ejection fraction and the recent measurements of weight,
blood pressure, and heart rate was based on both the role of that information in the
provider's cognitive work, but also on the HIT sociotechnical issue of how long it
takes providers to access that information by themselves in the EHR.
Increasing the use of β-blockers in qualifying patients is a difficult endeavor, and
will require more than a clinical reminder, however sophisticated it may be. Nonetheless,
this clinical reminder addresses many barriers to guideline adherence, and shows promise
for improving treatment decisions. It can make providers more familiar with and knowledgeable
about the guidelines and the benefit to their patients. It supports them in developing
skill and comfort with titration. It facilitates better utilization of clinical pharmacy.
This clinical reminder integrates a great deal of clinical data and treatment guideline
information compared with typical clinical reminders. In that sense, it supports the
creation of an accurate “situation model.”[58] This is important, given the complexity of the decision and treatment being addressed.
Typical reminders target screening activities. Few if any other clinical reminders
in the VA or other health care systems involve consideration of this level of clinical
data, nor pharmacological interventions of this complexity.
In these respects, this study is helping to explore the use of clinical reminders
and other CDSS tools for more complicated decision processes, and involving the integration
of more sophisticated algorithms (e.g., those using free text analysis). Furthermore,
this is being done in the relatively distributed context of primary care.
These challenges call for more proactive development methods. Our study shows the
feasibility and value of using CTA-based requirements elicitation to incorporate implementation
and adherence frameworks early on in the development process.
Limitations
Limitations of the Clinical Reminder Design
One limitation of our design is inherent with the clinical reminder function. The
provider must access that patient's record. If a provider does not have an appointment
with that patient or other reason to look in that record, the patient's treatment
will not be evaluated. A future tool we are planning is a periodic report providing
information on qualifying patients at risk, which can be integrated with panel management
tools.
One compromise in the design is the amount of clinical data it shows. There are a
very large number of potentially relevant factors for a clinician to consider for
the decision to initiate β-blocker titration. We could fit only a small number in
the design without overwhelming the use. We prioritized those data that would facilitate
quick decisions whether to exclude or continue evaluating. For many cases, providers
will need to look into the patient's record.
The announced change in the VHA's EHR to a new platform[59] presents a limitation in terms of the longevity of the specific design solution.
However, clinical reminder functions are common across all major EHR platforms.[60] The results of this study will inform the design of a new β-blocker treatment reminder
for the new platform, as well as for platforms used by other health care systems.
Furthermore, the broader findings of this study (that requirements development can
be aided though the use of CTA-type requirements elicitation enhanced via implementation
and sociotechnical frameworks) not only remain relevant, but also are arguably more
important now given the requirements development that must take place as part of customization
and configuration of the new platform.
Limitations of the Study
Our CTA relied only on interviews. We were unable to conduct observations of providers
working on HF treatment issues. We had providers go through case simulations, but
only as part of the preliminary evaluation of the prototype.
Another limitation is that our CTA was conducted to inform the design of the CDSS.
It was not designed to explore the full complexity of teamwork in the context of chronic
disease management. Additionally, our sample of PCPs for the CTA interviews included
only two physicians (the rest were nurse practitioners), potentially underrepresenting
physicians' views.
The tool has not been implemented yet, so we do not have any data on the impact to
guideline adherence or impact to patients care or resulting impact to health.
Conclusion
We have developed a decision support tool to address the lack of guideline adherent
β-blocker treatment for HF patients with depressed ejection fraction. To address the
challenges with implementation of assessment for and initiation of titration, we conducted
a CTA-based requirements elicitation process that incorporated three relevant frameworks
(PARIHS, HIT sociotechnical, and Cabana). The incorporation of these frameworks contributed
to design features that should facilitate providers' decision making, resource use,
and treatment execution.
Clinical Relevance Statement
Clinical Relevance Statement
Our work explores how clinical reminders could be used to help increase the rate of
qualifying heart failure patients (with depressed ejection fraction) who receive titration
to optimal doses of appropriate β-blockers. Our work also describes methods that should
reduce the risk of developing an HIT tool that does not fit into the organization.
Incorporating PARIHS, HIT socio-technical frameworks, or similar into the requirements
elicitation and development process can help identify needs regarding implementation
and integration into the larger sociotechnical system.
Multiple Choice Question
According to our study, one way to ensure a CDSS design incorporates factors related
to guideline adherence and implementation is to:
-
Include questions relevant to adherence and implementation in the requirements elicitation
phase.
-
Follow the design of previous CDSS tools that are in use.
-
Include mechanisms for penalizing providers who fail to follow the guidelines.
-
Make sure it is presented constantly to providers, with no way to quickly dismiss
it, to help embed it in memory.
Correct Answer: The correct answer is option a, “Include questions relevant to adherence and implementation
in the requirements elicitation and development phase.” Including such questions in
the requirements elicitation and development phase will help make sure that the design
addresses the needs related to why providers may not be following the guidelines,
and why providers may not use the CDSS tool. Answer B (“Follow the design of previous
CDSS tools that are in use”) may not result in a design that works for the specific
clinical decision at hand. Answers C and D are ways that may not actually help with
adoption of the tool or adherence to the guideline, and will likely have serious negative
impact on providers' performance and morale.
