Neopenda Case Study

Neopenda tasked our team to work on a 4-in-1 vitals monitoring system for neonates in low-resource environments.

Table of Contents

1. Project brief: We entered the world of Ugandan neonatal nurses.
2. Kickoff: We honed in on the needs of Ugandan nurses.
3. Data viz and current prototype: We investigated best practices in visualizing data in various domains, and checked out the current prototype.
4. User research: We stepped into the scrubs of healthcare providers in low-resource settings.
5. Synthesis: We identified emerging patterns.
6. Ideation and concepting: We generated ideas to tackle the issues of scanning vital signs and finding neonates in trouble. We made sense of contradictory insights between nurses and doctors.
7. Task flows: We took a step back to envision the platform in the context of nurses’ workflows.
8. Usability testing: We placed our prototype in front of healthcare providers.
9. Iteration and handoff: We iterated on our solution, and packaged it for further testing in Uganda.
10. Reflection: Out of the NICU, into the world.

My roles in this project:

• UX Designer/Lead Prototyper: In addition to conducting interviews, concept tests, and usability tests with my team, I prototyped the most critical feature of our solution in Axure over two days, delegated the converged prototype design to team members, and refined the final prototype after testing.
• Facilitator: I led and facilitated group activities for synthesis, ideation, information architecture, and goal-setting for internal group and external client meetings.
• Tech translator: I was familiar with the tech stack used by Neopenda’s dev team to create their web app prototype, including Node.js and React, and took the time to translate descriptions of technology to the design team.

We entered the world of Ugandan neonatal nurses.

Africa has the highest fertility rate in the world. However, the lack of infrastructure, skilled people, and functioning equipment results in bleak outcomes for newborns. Many newborns are premature and have problems at birth, which are exacerbated by the lack of monitoring equipment, and the disproportionate newborn-to-nurse ratio in NICUs. Our client, Neopenda, passed along several statistics comparing NICUs in Uganda to the United States. In low-resource environments, nurses have to manually check newborn vitals one at a time in NICUs with a 20:1 newborn-to-nurse ratio. In comparison, NICUs in the United States have a 1:1 newborn-to-nurse ratio and high-end equipment to monitor sick newborns.

Neopenda’s goal is to bring vital signs monitoring, a basic tenet of clinical care, to Uganda and other low-resource environments in an affordable device.

Neopenda’s headband device prototype

Our client:
Sona (CEO) and Tess (CTO) met at Columbia University completing their Master’s in Biomedical Engineering. They conducted interviews and design feedback sessions with over 150 nurses and doctors at 43 health facilities in every region of Uganda.

Our audience:
We focused on nurses and physicians (neonatologists, pediatricians, obstetricians) in Ugandan health facilities who work with newborns, with a focus on ill newborns in neonatal special care units (NICUs). In Uganda, neonatal nurses tend to be female, 20–50 years old, speak English, and educated in nursing, but low-income and may have limited familiarity with tablet/app technology.

It was exciting to work on a project that impacts the lives of neonates in low-resource environments. In addition to impacting the lives of Ugandan babies, I had the unique opportunity to develop a remote test plan to test our solution in Uganda, speak with Ugandan nurses and understand their unique perspectives, and face the challenge of designing an effective display of quantitative data. In addition to bringing a web development background and familiarity with the tech stack Neopenda’s development team used to build the initial prototype, I also brought my perspective of living in a developing country (Indonesia), seeing the chaos and disorganization of medical wards in countries outside the United States, as well as my experience speaking to non-native English speakers.

We honed in on the needs of Ugandan nurses.

After our initial kickoff meeting with Neopenda, we still had many unanswered questions, and we met with our clients again on the following day to continue our discussion and, more importantly, to align with them on the most important problem for their product to tackle.

We conducted a problem statement exercise with Neopenda. Tess and Sona whiteboarded problem statements for every possible Neopenda user. In the front is a weighted neonate model and headband prototype.

Tess and Sona mapped out the ecosystem the vital signs monitoring device lives in. We found that the stakeholders of their product include much more than nurses and physicians. They include:

• the neonates themselves
• their mothers
• hospital admins, e.g. Ugandan Ministry of Health officials involved in allocating budget towards staff and equipment
• physicians
• nurses

And they all had different wants and needs.

From this exercise, we defined the problem: nurses need a way to know which patient lives are in the most danger so that they can improve care, save lives, and be efficient, because they currently have too many tasks to manually take care of at once.

The most drastic difference in needs between stakeholders were between hospital admins, who prioritized budget and increasing efficiency over lives saved, physicians, who take a clinical approach and desire aggregated information on patients in order to diagnose them, and nurses, who are on the ground, doing most of the tasks to care for neonates, and are grossly understaffed.

When asked to highlight the most important user to focus on, Sona and Tess chose nurses, because nurses are the ones who have the ability to make the greatest impact on saving lives. We didn’t know it then, but this focus would enable us to sort through lots of contradictory insights in the coming weeks to reach our final solution.

We investigated best practices in visualizing data in various domains, and checked out the current prototype.

With our initial problem statement in hand, we moved on to conducting desk research and surveying the competitive landscape. For desk research, we wanted to gain a better understanding of the current state of neonatal vital signs monitoring systems, and address the challenge of clearly visualizing lots of quantitative information.

Figure 1: NICU display. For most of our concepts, we took a similar approach as existing NICU displays, using cards to display individual patient vital signs.

Figure 2 (left): stock market screens. Figure 3 (right): flight status screens. Inspired by stock tickers and flight status boards, we generated a concept utilizing a single-column layout of rows to display as many patient vitals as possible on a single screen.

I focused on conducting the competitive analysis with one teammate while the others focused on domain research. Neopenda already conducted their own competitive analysis of their direct competitors, existing vital signs monitoring devices within and outside of low-resource settings, and one teammate validated their findings. I wanted to investigate unexplored and related domains for inspiration, so I conducted the analysis of indirect competitors: EMT/first-responder software and livestock health monitoring devices. A key aspect of our problem specific to low-resource settings was dealing with the display of multiple patients at the same time rather than a single patient, which was the focus of elderly health monitoring devices, another possible indirect competitor. EMT/first-responder software helps users coordinate staff to handle urgent situations, and livestock health monitoring devices help farmers track and display the health of multiple animals on a single screen, both of which were relevant approaches to solving for our problem statement.

Figure 4 (left): Intrepid Response, EMT software which features a map that auto-navigates to selected users and user-generated points of interests. Later, our team tested a map concept inspired by this indirect competitor. Figure 5 (right): Herddogg offers livestock health monitoring devices and a dashboard web/mobile app.

We also conducted a heuristic evaluation of the existing product, and I conducted the final analysis of our evaluation for our client. I found that most usability heuristic violations occurred on the dashboard screen and patient information view/edit screens.

On the dashboard, usability heuristics error prevention, recognition over recall, and aesthetic and minimalist design were most violated. On the patient information view/edit screens, usability heuristics match between system and real world, and recognize and recover from errors were most violated.

We stepped into the scrubs of healthcare providers in low-resource settings.

In addition to understanding the display of vast quantities of information through domain research, assessing the competitive landscape, and evaluating the usability of the current dashboard prototype, our team wanted to be purposeful in our interview questions as we jumped into user and SME research.

Neopenda provided us with users and SMEs with a diverse range of experiences working with neonates. I analyzed our interviewees’ job titles and background descriptions and created a 2x2 matrix for our team to organize our research questions and interview scripts by audience types:

• users who work in close proximity to neonates, e.g. “on-the-ground” nurses, vs. physicians with a clinical perspective.
• users with frequent exposure to low-resource environments like Uganda, vs. less-to-no exposure to low-resource environments.

We wanted to learn the following from our interviews:

• NICU environments in Uganda
• Differences between nurse and doctor roles at the NICU, and specifically whether or not nurses should have the same permissions as doctors to update alarm parameters
• Nurses’ processes and ways they currently respond to NICU emergencies
• Initial user reactions to the existing web app prototype, which was a specific ask from Neopenda

We remotely interviewed a total of 7 nurses and doctors — 5 SMEs and 2 users across various time zones — over Zoom video conference and WhatsApp phone calls. In addition to the poor network connection during our international remote interviews, my team faced language barriers when attempting to speak with our Ugandan users. It was a surprise that while users spoke English, they had a heavy accent and lower English speaking level, and we had to learn on the fly how to adjust interview questions to increase comprehension by non-native English speakers. After our first call with a Ugandan user, my team and I approached future interviews with a better understanding of subtle language nuances when speaking with Ugandans.

We conducted all our interviews remotely with users/SMEs in Uganda, Zambia, and several U.S. states.

We identified emerging patterns.

After conducting interviews, we needed to make sense of our data, and decided to synthesize results. We had many contradictory interview insights, and keeping track of each interviewee’s background would help us prioritize data from users closest to Neopenda’s target audience; I advocated for increasing specificity of the categories we used, by including a separate color for quotes, and coding the perspectives each data point came from: Ugandan vs. U.S. nurses, Ugandan vs. US doctors. In addition to facilitating the affinity diagramming and conducting a voting exercise on the top 5 categories to focus on, I took notes on our follow-up discussion on each team member’s reasoning behind specific category votes.

I voted for “patient records” as an area of focus, because I wanted to keep in mind the most valuable aspects of the platform for stakeholders who would have the most impact in financial decisions — officials and administrative staff.

For the Ministry of Health and administrative staff, keeping track of patient records is a huge value proposition of the product, since it influences budgetary decisions and can increase efficiency. Our team unanimously voted for “physical location”, because it was an important issue for our platform to address in order for the product to be a success. “Visual display” seemed crucial to providing the “physical location” of babies, so our team also voted unanimously for that category. By voting on the top 5 categories to focus on, our team gained alignment on the 4 specific issues we wanted to address in our solution.

We saw several emerging trends, and as a group identified four categories of focus: (1) visual display, (2) alarms, (3) patient records, and (4) physical location of babies.

In summary, here’s what we found from our research:

Things that weren’t working with the existing prototype:

• Patient vital signs were difficult to scan on the screen.
• Bed labeling for each patient’s vital signs card didn’t make sense to Ugandan nurses, which indicated a mismatch between the system and the real world. Many cots, incubators, and phototherapy beds are shared between multiple patients.

Things that surprised us about the neonatal monitoring workflow and process:

• A Ugandan doctor admitted that nurses should have power and access to change alarm parameters, because sometimes they are more familiar with equipment and the patients.

“Nurses know equipment better, and should be able to change the alarms” — Ugandan doctor

• Contradictions between what visitors vs. natives of Uganda said regarding frequency of and access to patient records.
• Lack of organization in health facilities. This was shocking to me, and in my eyes indicative of the most important issue to address: locating babies in a disorganized ward.

“For each baby admitted, they have a form filled on top of the baby.” — Ugandan nurse

“Babies [are] all over. The midwives and nurses have to move around, looking baby by baby [at] the label.” — Ugandan nurse

After synthesizing these insights, we revisited and clarified our problem statement:

There is currently a lack of medical equipment, space, and staff in low-resource environments. Neonatal nurses in low-resource environments need a way to efficiently scan vital monitoring information and find neonates in critical condition so that they can channel their energy to the child with most need.

I synthesized our research into ideas for guiding principles to support our problem statement, which our team ended up choosing from and naming as a group.

We presented our findings to Neopenda, and by the end, Tess mentioned that we “were able to scope in on the core issue of the problem.” With efficient vital signs scanning and neonatal finding at the forefront of our minds, our team set off to start ideating and concepting.

We generated ideas to tackle the issues of scanning vital signs and finding neonates in trouble.

I led several brainstorming exercises, and we each took a stab at sketching concepts out on paper. After presenting concepts to each other, we combined similar concepts and added a few more divergent ideas.

Revisiting and combining concept ideas into prototypes for concept testing.

We started with 5 SMEs/users and 2 Ugandan nurses scheduled for concept testing, but one Ugandan nurse fell through, and the other had technical difficulties in maintaining an international call. In the end, we tested our concepts with 5 SMEs/users total, and converted our remaining Ugandan nurse concept test into an additional user interview, where we gained insights into the neonatal admissions process.

C1: Moving map

We removed C1 after two tests due to poor testing results.

C2: Queue up

We kept this concept in play based on testing results.

C3: Don’t snooze (my concept)

We kept this concept in play based on testing results.

C4: Dynamic list

We kept this concept in play based on testing results.

C5: Switchboard

We kept this concept in play based on testing results.

In summary, from our concept tests, we learned the following:

• Nurses and doctors have different needs and wants for a vital signs monitoring solution. Some features are more desirable to doctors than nurses, e.g. trends visualizations, nurse notes, alarm counters.
• Our solution should not create additional work for nurses. We removed several concept parts, e.g. C1’s map layout of patients, C2’s nurse notes, C5’s baby location drop-down fields. We removed C1: moving map, because it would create additional work to organize the layout of babies.
• Nurses respond to and prioritize different vital sign alarms in different ways, so it’s important to indicate which vital sign is currently abnormal, and use a different color to differentiate temperature from other vital signs. Users also suggested being reminded every hour instead of having constant alarms for abnormal temperatures.
• We gained a better grasp of our users’ mental models, and updated our patient information screen accordingly.

We made sense of contradictory insights between nurses and doctors.

I advocated for each team member to revisit one concept tester’s interview notes so that we could later return as a group to orally and visually synthesize results by concept. Everyone spent some time empathizing with their user, and then, per concept, everyone gave their input from the perspective of their tester, which we wrote down on the whiteboard.

Then, with our concept testing insights in hand, we returned to the drawing board to collectively imagine a converged prototype which coherently pieced together the user-desired parts of our 4 remaining concepts.

Our team deliberated on how to make sense of the contradictory insights, especially between nurses and doctors. C4 was particularly divisive, and after much deliberation, we decided to do the following:

1. Prioritize the viewpoints of nurses, especially if they worked closely to or are Ugandan nurses.
2. Take what users appreciated most out of C4, the comprehensive list of patient statuses, and combine it with C5, the ever-present view of patient alarms, to create a solution that helps nurses scan vital signs data and find neonates in the most efficient manner possible.

When we presented our findings and the challenges we faced when making sense of contradictory insights, our client advised us to “go with the best decision possible” moving forward, and prioritize nurses’ perspectives.

We took a step back to envision the platform in the context of nurses’ workflows.

At this point in the project, I felt that we lacked a comprehensive understanding of the whole system, and needed to take a step back to map out nurse workflows using our future Neopenda dashboard solution before deciding on specific tasks to prototype and test. So I led our task flow whiteboarding exercise, and systematically walked through each step nurses would take during their workflow, noting 4 specific action types:

1. Nurse actions: caring for neonates, interacting with the dashboard
2. Events in the hospital: emergencies, abnormal vital signs
3. Physical headband device states and actions
4. Dashboard states and actions

We mapped out 3 task flows: (1) add new patient, (2) monitor patient, and (3) discharge patient.

Figure 6 (left): Task flow key. Figure 7 (right): Task flow 1, Add new patient.

Figure 8 (left): Task flow 2, Monitor patient. Figure 9 (right): Task flow 3, Discharge patient.

From our task flow mapping exercise, we found the following opportunities to update our prototype IA and UX:

• Allow users to assign both new and existing patients to devices in two specific situations: (1) initial device connection, (2) device recharge and replacement. This finding informed the links and content we included in our prototype, specifically the Patients and Devices screens.
• Include a patient archival user flow and interface.

With our task flow insights in hand, we decided to delegate the prototype. Due to our time constraints, I built the most critical and complicated piece of the prototype — the vitals dashboard with the auto-scrolling alarms sidebar and vital signs cards featuring colored alarms — since my web development background would help us efficiently blast through Axure prototyping.

I built out components of our Axure team style guide, the dashboard screens and interactions, as well as tied the prototype together by incorporating my team’s modals and navigation links.

We placed our prototype in front of healthcare providers.

We had four main goals with usability testing:

1. Understand if the updated dashboard helps users scan patient vitals
2. Assess if the summarized alarm status bar on the left side of the screen helps or hinders the user’s workflow. This was of particular interest to us, since the prototype included untested ideas based on concept testing synthesis.
3. Determine how users respond to separate temperature alerts, which we differentiated using color: blue for low temperatures, red for all other vital sign alerts. This was also a new idea we decided to prototype and test.
4. Iterate on our solution based on user feedback.

We conducted usability tests with 4 SMEs: 1 in-person test with a neonatal physician and 3 video conference calls with nurses. Our team visited a NICU in Chicago for our in-person test, and saw a calm and quiet ward. The team had to talk in quiet voices, which we imagine starkly contrasts the experience in Uganda.

We faced many technical and logistical challenges in our usability tests. Our prospective Ugandan tester lacked internet access, a computer to test the prototype on, and video chat capabilities. While we did our best to work with Neopenda’s Ugandan coordinator who had internet and computer access, the coordinator was available for only limited times; the Ugandan nurse missed our scheduled test time and had additional scheduling conflicts.

We didn’t test our mid-fidelity prototype with native Ugandan nurses, and developed a research plan to hand off to Ugandan test coordinators after our project ended.

We put our prototype in front of healthcare providers, and gave them 4 tasks informed by our workflow exercise.

Task 1: Monitor vitals

Task 2: Add patient and set alarm parameters

Task 3: View trends

Task 4: Discharge patient

After testing, we immediately made updates to our prototype before preparing for a final presentation of our findings:

• We updated “Archive” to say “Remove” on the Patients screen to match users’ mental models.
• I combed through our final prototype for font size, font weight, color, spacing, and other style inconsistencies, and updated labeling on the Dashboard screen based on user feedback.

We iterated on our solution, and packaged it for further testing in Uganda.

In addition to prototype updates, I worked on the following parts of our final handoff:

• Designed task flows for presentation to client.
• Advised teammate in sitemap structure and labeling, pulling in my web development background.
• Developed a team template in Sketch for annotated wireframes, then annotated the dashboard screens and interactions I developed.
• Developed a UI kit in Axure for Neopenda.
• Research plan goals and resources, including an ideal path walkthrough video and testing setup instructions.

While we provided a unique solution to Neopenda which solved specifically for the needs of Ugandan neonatal nurses in low-resource settings, we still had many unanswered questions by the end of the project. Given more time and resources, I would love to put our solution in front of native Ugandan users, specifically to validate the updated dashboard interface design.

Neopenda’s next steps will be implementation of the native Android tablet prototype, a clinical trial of the headband device with Tufts Medical Center, and in-country testing of our solution with the help of coordinators in Uganda.

Our solution: mid-fidelity Axure prototype.

We outlined 5 recommendations for Neopenda to implement in their native Android prototype, prioritized by time urgency:

In addition to providing future recommendations for the vitals monitoring dashboard, our team packaged a test plan with resources for Neopenda coordinators to test with native Ugandan users. After discussing test plan ideas with the team, I synthesized our test plan goals:

1. Create something easy to iterate upon. Our test plan needed to be flexible and modular, so that it can evolve as Neopenda gains new user insights and further develops their product.
2. Create something easy for test teams new to UX. Our test plan needed to prepare the Ugandan team for three phases of testing: setup, conducting the test, and analysis of results.
3. Setup: I wrote setup instructions for testing the Axure prototype on Android tablets using either the Axshare app or the exported Axure prototype files. I also included resources for recording usability test videos and/or audio on the computer, tablet, and phone.
4. Conducting: In addition to the test plan, we included an ideal test path walkthrough video, so that moderators understand the assumed route users will take as they complete tasks. I planned and recorded this video for our client. We also included instructional videos and resources with best practices for conducting tests, i.e. asking open-ended questions, assigning notetaker/recorder roles.
5. Analysis: We provided a spreadsheet to collect, organize, and analyze test results, with examples of scenarios, tasks, outcomes, and recommendations.
6. Easy for ESL test teams, e.g. including a vocabulary list/legend for UX terminology.
7. Test the prototype with primary users, Ugandan nurses, to see the overall usefulness of the solution, how it fits into their workflow, and test anything that required further testing marked in our annotated wireframes.

Out of the NICU, into the world.

Through working with Neopenda, I gained confidence in my process as a UX designer. I started my journey in Designation as an awkward developer comfortable talking to a room of professional or aspiring developers, but not to a wider audience of designers, business developers, and clients. Now I feel comfortable presenting to a variety of audiences, designing collaborative client meeting exercises, and applying my goal-oriented mindset to client meetings.

From my client projects at Designation, I discovered my love of whiteboarding and making sense out of chaos through user research synthesis and site/user flow mapping. Working with Neopenda was immensely gratifying, because the project was meaningful as a product that saves infant lives, and I thoroughly enjoyed applying my synthesis, concepting, and information architecture skills. Through previous client projects, I learned how to reconcile contradictory user insights, especially during concept testing. With Neopenda, I learned how to use client feedback and a specific user focus to transform conflicting preferences and insights into a solution that helps NICU nurses in low-resource settings save lives.

I also realized that I love paying attention to the details, and am driven to comb through deliverables and prototypes at the end for visual and functional consistency.

In addition to doing what I love, I picked up valuable skills in this project:

• I learned how to coordinate and conduct remote interviews, concept tests, and usability tests with an international audience, including ESL users.
• I did a fast, deep dive into neonatal care. I gained knowledge of neonatal care in the U.S. and low-resource settings such as Uganda, from both nurses and physicians.
• I developed materials to pass off to non-design-savvy professionals. I learned how to develop an out-of-country usability test plan to pass off to Ugandan test coordinators.