Presentation
User-centered Design of Applications Embedded in Electronic Health Records: Case Study of a Visualization Application for Heart Failure
Description*Background:
Despite the prevalence of electronic health records (EHRs), EHR users frequently report being hindered by documentation issues such as incomplete documentation or lack of standardization (e.g., Ni et al., 2021). This particularly affects clinicians who manage complex medication regimens, such as cardiologists caring for patients with heart failure with reduced ejection fraction (HFrEF). HFrEF is best managed by guideline-directed medical therapy (GDMT), a set of medication titration guidelines (Maddox et al., 2021). While GDMT is effective, its clinical compliance is low due to a lack of standardized documentation (Greene et al., 2018). While there have been attempts to address low GDMT compliance (e.g., Spaulding et al., 2023), efforts to improve GDMT documentation within EHRs have been limited or undocumented.
User-centered design (UCD) is a methodology that has been used to solve complex healthcare problems, including GDMT-related ones (Spaulding et al., 2023). UCD consists of stakeholder involvements and repeated cycles of research, design, and evaluation; and is suited for projects with complex dependencies like EHRs. Using UCD, we designed an EHR-integrated tool to improve GDMT documentation. We recruited patient, clinician, and technologist stakeholders from a healthcare system located in the Midwestern United States. We identified barriers faced by clinicians making GDMT decisions based on EHR documentation and then designed and evaluated prototype solutions with stakeholders working in a heart failure clinic.
We implemented five strategies not commonly documented in the context of EHR application design: unconventional designs in early design stages, personas for evaluation, short scales to collect usability and acceptance feedback, a technology taskforce, and a patient and family advisory council. For each strategy, we present our experience, reflections considering current literature, and design implications for researchers and practitioners who want to employ similar strategies in their work.
• Unconventional Designs:
We designed unconventional visualizations to direct users to look at important information for GDMT decision-making through metaphors. Examples of visualizations included pill bottles representing a patient’s current medication dose relative to the guidelines and a tree displaying data points relevant for GDMT decision-making as leaves. Most stakeholders liked these unconventional designs and reported how the visuals sped up patient information gathering. Clinicians favored simpler designs (e.g., pill bottles) over more complex designs (e.g., tree).
Our design ideas presented GDMT information graphically which helped with information retrieval according to stakeholders. Visually busier designs like the tree were less preferable as our participants perceived them as more difficult to interpret. Some design elements also contradicted EHR design guidelines, e.g., clicking on tree branches to display information instead of direct access to data with limited mouse interaction (Zahabi et al., 2015).
DESIGN IMPLICATION: Unconventional designs such as visualizations relying on metaphors can reduce information density while drawing users’ attention to specific data points. However, unconventional designs must also conform to existing guidelines to ensure a consistent user experience.
• Personas for Evaluation:
Two research team members – registered nurses with a cardiology experience – created both simple (e.g., recently diagnosed with HFrEF with no prior history) and complex personas (e.g., with a cardiac implantable electronic device and comorbidities) for testing purposes. Each persona had a 6-month visit history, including clinical decisions and medication adjustments. In earlier evaluation stages, we presented prototypes based on simple personas to focus on the GDMT tool’s purpose and goals. In later stages, we used complex personas to pressure test the prototype and identify where the tool might fail in supporting clinicians effectively.
While requiring a significant amount of time to develop, the fidelity of the personas fostered consistency between prototypes and helped clinicians stay focused on evaluating the prototypes and not the personas.
DESIGN IMPLICATION: Personas are essential for both the design and the evaluation of EHR tools and should be crafted carefully by people with clinical expertise.
• Short, Standardized Scales:
As our evaluation sessions with clinicians were limited to 30 minutes (representative of the duration of a patient appointment), we chose two short scales, UMUX-LITE (2 items) and Net Promoter Score (NPS; 1 item), to gather consistent usability and user experience metrics on three prototypes that emerged from early UCD rounds. We administered these scales for each prototype, resulting in responses to 9 standardized instrument items per participant.
While we could not use longer but more robust instruments (e.g., SUS), the shorter scales enabled us to ask more interview questions to contextualize the scores clinicians gave. This approach helped us discard lower ranking prototypes efficiently.
DESIGN IMPLICATION: Short instruments like UMUX-LITE and NPS provide reliable data to help choose between designs.
• Technology Taskforce:
We formed a 10-people interdisciplinary technology taskforce with data engineers and EHR experts. We first met with the taskforce quarterly then switched to meeting taskforce members individually as needed. Topics discussed included best implementation pathways sanctioned by the EHR manufacturer and location of GDMT-related data points within EHR databases.
Effective strategies for research teams to collaborate with technological teams on UCD projects are crucial but have been seldom documented for hospital applications. A few strategies we implemented, such as having a large taskforce, did not prove successful as some key team members did not participate despite being the only ones to hold knowledge on specific topics such as data location.
DESIGN IMPLICATION: Bringing in EHR and data experts is essential to implement ideas that are not just appealing but also feasible. While large groups can be useful in earlier UCD cycles, consider individual meetings and task assignments if key participants are not contributing.
• Patient and Family Advisory Council:
We formed a Patient and Family Advisory Council (PFAC) with 9 HFrEF patients and family members. The council met twice to solicit their feedback on prototypes and share their lived experiences as a HFrEF patient or caregiver. Demonstration of the prototypes and general discussion illuminated the challenge of taking multiple prescription medications and the complexities of receiving care from multiple providers. After each meeting, we derived questions to ask our clinical stakeholders, e.g., “should the tool integrate cost-sensitive prescribing?”
PFAC participants raised issues that the research team had not considered in initial UCD rounds. Many insights derived from the PFAC meetings led to low-priority requirements, understandable, as clinicians—not patients—are the tool’s primary users. Levels of engagement also varied among participants; some were intrigued by the prototypes and suggested improvements; others appeared indifferent.
DESIGN IMPLICATION: The voice of HFrEF patients and their caregivers is valuable for clinician-facing technology design as their needs can inform design decisions, e.g., sensitive cost prescribing. However, we suggest prioritizing clinician involvement in earlier UCD cycles and involving patients in later cycles.
*Future Work:
We plan on moving from prototyping to development and integration of the GDMT tool within our EHR, measuring its impact on GDMT compliance and EHR documentation.
*References:
• Greene, S., …Fonarow, G. (2018). Medical Therapy for Heart Failure With Reduced Ejection Fraction. JACC, 72(4), 351-366.
• Maddox, T., …Youmans, Q. (2021). 2021 Update to the 2017 ACC Expert Consensus Decision Pathway for Optimization of Heart Failure Treatment…. JACC, 77(6), 772-810.
• Ni, Y., …Beal, S. (2021). Automated detection of substance use information from electronic health records for a pediatric population. JAMIA, 28(10), 2116-2127.
• Spaulding, E., …Martin, S. (2023). Using Human-Centered Design Methodology to Identify Challenges and Inform the Development of a Digital Toolkit to Optimize Heart Failure Guideline-Directed Medical Therapy From Diverse Clinician, Patient, and Patient Health Partner Perspectives. Circulation, 147(Suppl_1), AP398-AP398.
• Zahabi, M., Kaber, D. , & Swangnetr, M. (2015). Usability and Safety in Electronic Medical Records Interface Design: A Review of Recent Literature and Guideline Formulation. Human Factors, 57(5), 805-834.
Despite the prevalence of electronic health records (EHRs), EHR users frequently report being hindered by documentation issues such as incomplete documentation or lack of standardization (e.g., Ni et al., 2021). This particularly affects clinicians who manage complex medication regimens, such as cardiologists caring for patients with heart failure with reduced ejection fraction (HFrEF). HFrEF is best managed by guideline-directed medical therapy (GDMT), a set of medication titration guidelines (Maddox et al., 2021). While GDMT is effective, its clinical compliance is low due to a lack of standardized documentation (Greene et al., 2018). While there have been attempts to address low GDMT compliance (e.g., Spaulding et al., 2023), efforts to improve GDMT documentation within EHRs have been limited or undocumented.
User-centered design (UCD) is a methodology that has been used to solve complex healthcare problems, including GDMT-related ones (Spaulding et al., 2023). UCD consists of stakeholder involvements and repeated cycles of research, design, and evaluation; and is suited for projects with complex dependencies like EHRs. Using UCD, we designed an EHR-integrated tool to improve GDMT documentation. We recruited patient, clinician, and technologist stakeholders from a healthcare system located in the Midwestern United States. We identified barriers faced by clinicians making GDMT decisions based on EHR documentation and then designed and evaluated prototype solutions with stakeholders working in a heart failure clinic.
We implemented five strategies not commonly documented in the context of EHR application design: unconventional designs in early design stages, personas for evaluation, short scales to collect usability and acceptance feedback, a technology taskforce, and a patient and family advisory council. For each strategy, we present our experience, reflections considering current literature, and design implications for researchers and practitioners who want to employ similar strategies in their work.
• Unconventional Designs:
We designed unconventional visualizations to direct users to look at important information for GDMT decision-making through metaphors. Examples of visualizations included pill bottles representing a patient’s current medication dose relative to the guidelines and a tree displaying data points relevant for GDMT decision-making as leaves. Most stakeholders liked these unconventional designs and reported how the visuals sped up patient information gathering. Clinicians favored simpler designs (e.g., pill bottles) over more complex designs (e.g., tree).
Our design ideas presented GDMT information graphically which helped with information retrieval according to stakeholders. Visually busier designs like the tree were less preferable as our participants perceived them as more difficult to interpret. Some design elements also contradicted EHR design guidelines, e.g., clicking on tree branches to display information instead of direct access to data with limited mouse interaction (Zahabi et al., 2015).
DESIGN IMPLICATION: Unconventional designs such as visualizations relying on metaphors can reduce information density while drawing users’ attention to specific data points. However, unconventional designs must also conform to existing guidelines to ensure a consistent user experience.
• Personas for Evaluation:
Two research team members – registered nurses with a cardiology experience – created both simple (e.g., recently diagnosed with HFrEF with no prior history) and complex personas (e.g., with a cardiac implantable electronic device and comorbidities) for testing purposes. Each persona had a 6-month visit history, including clinical decisions and medication adjustments. In earlier evaluation stages, we presented prototypes based on simple personas to focus on the GDMT tool’s purpose and goals. In later stages, we used complex personas to pressure test the prototype and identify where the tool might fail in supporting clinicians effectively.
While requiring a significant amount of time to develop, the fidelity of the personas fostered consistency between prototypes and helped clinicians stay focused on evaluating the prototypes and not the personas.
DESIGN IMPLICATION: Personas are essential for both the design and the evaluation of EHR tools and should be crafted carefully by people with clinical expertise.
• Short, Standardized Scales:
As our evaluation sessions with clinicians were limited to 30 minutes (representative of the duration of a patient appointment), we chose two short scales, UMUX-LITE (2 items) and Net Promoter Score (NPS; 1 item), to gather consistent usability and user experience metrics on three prototypes that emerged from early UCD rounds. We administered these scales for each prototype, resulting in responses to 9 standardized instrument items per participant.
While we could not use longer but more robust instruments (e.g., SUS), the shorter scales enabled us to ask more interview questions to contextualize the scores clinicians gave. This approach helped us discard lower ranking prototypes efficiently.
DESIGN IMPLICATION: Short instruments like UMUX-LITE and NPS provide reliable data to help choose between designs.
• Technology Taskforce:
We formed a 10-people interdisciplinary technology taskforce with data engineers and EHR experts. We first met with the taskforce quarterly then switched to meeting taskforce members individually as needed. Topics discussed included best implementation pathways sanctioned by the EHR manufacturer and location of GDMT-related data points within EHR databases.
Effective strategies for research teams to collaborate with technological teams on UCD projects are crucial but have been seldom documented for hospital applications. A few strategies we implemented, such as having a large taskforce, did not prove successful as some key team members did not participate despite being the only ones to hold knowledge on specific topics such as data location.
DESIGN IMPLICATION: Bringing in EHR and data experts is essential to implement ideas that are not just appealing but also feasible. While large groups can be useful in earlier UCD cycles, consider individual meetings and task assignments if key participants are not contributing.
• Patient and Family Advisory Council:
We formed a Patient and Family Advisory Council (PFAC) with 9 HFrEF patients and family members. The council met twice to solicit their feedback on prototypes and share their lived experiences as a HFrEF patient or caregiver. Demonstration of the prototypes and general discussion illuminated the challenge of taking multiple prescription medications and the complexities of receiving care from multiple providers. After each meeting, we derived questions to ask our clinical stakeholders, e.g., “should the tool integrate cost-sensitive prescribing?”
PFAC participants raised issues that the research team had not considered in initial UCD rounds. Many insights derived from the PFAC meetings led to low-priority requirements, understandable, as clinicians—not patients—are the tool’s primary users. Levels of engagement also varied among participants; some were intrigued by the prototypes and suggested improvements; others appeared indifferent.
DESIGN IMPLICATION: The voice of HFrEF patients and their caregivers is valuable for clinician-facing technology design as their needs can inform design decisions, e.g., sensitive cost prescribing. However, we suggest prioritizing clinician involvement in earlier UCD cycles and involving patients in later cycles.
*Future Work:
We plan on moving from prototyping to development and integration of the GDMT tool within our EHR, measuring its impact on GDMT compliance and EHR documentation.
*References:
• Greene, S., …Fonarow, G. (2018). Medical Therapy for Heart Failure With Reduced Ejection Fraction. JACC, 72(4), 351-366.
• Maddox, T., …Youmans, Q. (2021). 2021 Update to the 2017 ACC Expert Consensus Decision Pathway for Optimization of Heart Failure Treatment…. JACC, 77(6), 772-810.
• Ni, Y., …Beal, S. (2021). Automated detection of substance use information from electronic health records for a pediatric population. JAMIA, 28(10), 2116-2127.
• Spaulding, E., …Martin, S. (2023). Using Human-Centered Design Methodology to Identify Challenges and Inform the Development of a Digital Toolkit to Optimize Heart Failure Guideline-Directed Medical Therapy From Diverse Clinician, Patient, and Patient Health Partner Perspectives. Circulation, 147(Suppl_1), AP398-AP398.
• Zahabi, M., Kaber, D. , & Swangnetr, M. (2015). Usability and Safety in Electronic Medical Records Interface Design: A Review of Recent Literature and Guideline Formulation. Human Factors, 57(5), 805-834.
Event Type
Oral Presentations
TimeTuesday, March 261:50pm - 2:10pm CDT
LocationSalon A-1
Hospital Environments