Seattle Children's Hospital
a video preview of a banking app
Recap PCA is a proof-of-concept hardware and software prototype designed for outpatient opioid management.
Recap PCA was developed with the guidance of anesthesiologists at Seattle Children's Hospital / University of Washington. It is protected as a University of Washington CoMotion Invention #48371.
Product Designer
Primary Responsibilities
Lit. Reviews, Interface Design and UX, Hardware & Software Engineering
Mackenna Lees→(Comp. Analysis, Interface Design and UX, Content Design, Testing)
Scott Smith→ (Project Management, UX, Analytics, Industrial Design)
Jill Aneri Shah→ (Research Coordination, UX, Hardware Development, Usability)
After surgery, patients in a hospital setting are closely monitored and often given powerful opioid-based analgesics (painkillers). In some cases, patient-controlled analgesia (PCA) devices are used, allowing for pain management that is more tailored to the patient's specific experience, but are notably still monitored by healthcare professionals.

However, when patients move back home, they are often given a generous prescription for powerful yet addictive opioid-based analgesics (painkillers). Unlike the experience in a controlled hospital environment, patients are suddenly faced with the need to regulate their own pain medication while they recover. The combination of over-prescription, ineffective self-regulation, and a lack of awareness of drug return practices has contributed to an opioid crisis that has claimed more than 42,000 deaths and two million cases of opioid abuse in 2016 alone.

This project is an investigation into bringing the concept of patient-controlled analgesia into the home. Ultimately, this project aims to address, at least in part:
  • Lack of education and the resultant anxiety around proper management of pain at home
  • Historically overgenerous prescribing practices of opioids
  • Dosage vs. relative effect of a given painkiller is different for each patient; current one-size-fits-all approach does not accurately account for these difference
  • Lack of visibility into a patient's actual use of the opioids
  • Lack of ability to accurately step a patient down from opioids without significant confidence from the patient or time-intensive guidance from the doctor,
  • Low return rate of unusued opioids – unused opioids may abused by non-patients who find them in the home or may be illicitly sold
Research & Lit. Review
Competitive Analysis
We conducted a competitive analysis of twelve existing pill adherence, reminder systems, and dispensing products on the market to identify existing modalities, features, storage methods, and scenarios where these devices could be used. We found that many at-home devices on the market primarily serve one purpose: long-term, static-schedule pill adherence, especially for multi-pill situations such as morning vitamins and medication.

Contrastingly, the use-case for Recap PCA would be short-term management of controlled substances, none of which the competitors addressed.
User Interviews
To gain a better overall understanding of the problem space and current experiences from a user/patient perspective, we conducted seven semi-structured interviews, all the participants of which were  given opioids after a surgery, with ages ranging from 21 to 82 and representing a wide variety of backgrounds including students, medical professionals, and schoolteachers.
Literature Review
A literature review on a total of nine (9) background papers was conducted to gain a deeper academic and in-practice understanding of the problem, its implications, and existing interaction design conventions and existing medical devices on the market.

Some key quotes:
Overall Research Findings
Key high-level findings from our research included:
Product Direction
Through this research, we proposed our high-level solution: a connected, in-home opioid dispenser, which could gently step users down towards lower or less frequent doses of medication and provide caregivers with the resultant data.

This approach was approved by our stakeholders, and address a number of issues:
  • Being able to guide and monitor a patient with their own reported experiences (dose vs. individual experience of pain) allows healthcare professionals to better tailor prescribing practices to individual experiences and help prevent abuse
  • Automatically step down a patient from opioids to non-opioid painkillers to help prevent addiction
  • Preserving the agency of the patient, and still allow for use of opioids when necessary.
  • Patients having to report their pain levels regularly increases mindfulness of their dosing regimen.
  • Connected device surrounding use and dosing data can provide robust data to inform broader dosing practices of opioids
  • Return of hardware at end of recovery can improve the return rate of unused opioids, rather than having unused opioids lay around the patient home
A high-level overview of where the product fits within a typical use-case is outlined below:
Form and Hardware
Our team began to develop concepts for the hardware design of the device itself. Over the course of several weeks, we ideated on our own and reviewed them with the team (a sort of rapid design-only sprint), refining and sharing ideas around both the form and the internal dispensing mechanism itself.

We physically 3D-printed more than 25 different versions of internal dispensing components, form factors, and other ideas.
a video preview of a banking app
Ideation on form factors

Top-left: dispensing mechanisms, circular/cylindrical designs, “Nespresso”-inspired pill dispenser; top-right: notification ideas, integration of pill bottles; bottom-left: inspired by a coffee machine, integrating a handle, locking mechanism, hand-accomodating pill dispenser; bottom-center: pill bottles, machines, cup designs, novel designs; bottom-right: interface integration, form factors, integration with dispensing mechanism

a video preview of a banking app
Ideation on dispensing mechanisms

Left: exploded diagram of an upright circular mechanism, integration into a form; center: Circular-feeding, pill-bottle based, etc.; right: Various dispensing mechanisms, including circular, linear-feeding, and spring-loaded

User Interface and User Flow
In addition to the design of the physical device itself, an obvious key component to such a device is how a user might interact with it. We explored ideas relating to physical buttons, interfaceless devices, and integration into a patient's home environment, but settled on a small 3.5" touchscreen as more "low-tech" approaches were not able to fully address the fidelity of actions that were desired.

The user flow was based on three core concepts: self-reporting of pain level (via a numeric pain scale), a resultant recommended dose type / amount, and final selection by the patient of the actual medication to dispense.

A key part of the experience outside of the "golden path" of just collecting data, recommending, and dispensing is the number of secondary cases that relate to improper opioid use (not safe to ingest n amount of medication within a given timeframe) and patient agency (a patient should never feel as if they are truly unable to dispense medication if they really need it).
a video preview of a banking app
Preliminary user flow (UI ver. 1)

A preliminary, recommendation-based user flow was developed and tested with stakeholders, creating a flow that not only completed the desired task, but navigated existing power structures in a medical environment.  

Content Design
At its core, the device is acting as an in-home "replacement" for the guidance usually given by a healthcare professional. As patients may be experience anxiety, fatigue, and of course pain, the content and writing style we use is essential to making a user feel comfortable, in control at all times (esp. in emergencies), and hopefully as if the device was acting in the best interests or at minimum, neutral in relation to the patient's experience. While we had basic ideas of what this type of tone would be in our initial ideation and early drafts, the following usability testing would be integral to getting things right.
Analytics Framework
Collecting data on users of any device or service is not a benign action. Data may be stolen or abused with little regard for its impact on others. As a result, our team has carefully considered the role data collection plays in this device. Currently, opioid prescribing practices are based largely upon guesswork. Data collection in aggregate form can potentially play an important role in providing researchers and caregivers numeric data for prescription use, patterns, and behaviours. In light of this, we have determined that collecting such analytics as a device feature is worthwhile and aligns with both patient and caregiver values.
a video preview of a banking app
Rough idea of upper and lower bounds of pain at any given time

Device would be able to inform such a graph, to check for significant deviation from the “norm” and identify erroneous incidences of pain and/or pill dispensing

Usability Testing
Testing was conducted in an in-person moderated format. Two scenarios, a semi-structured interview protocol, and hypotheses were developed to test golden-paths and core functionality. Ten users of varying ages, backgrounds, and experiences with pain participated and signed consent forms for this moderated usability study.
a video preview of a banking app
Prototype UI for Usability Testing (UI approx. ver. 4)

This UI was tested with users on iPad via a Figma prototype. To keep manufacturing costs low, we initially proposed a UI with a non-touch screen navigated by seven buttons (X, low, med, high, no, yes, and override) but quickly found in testing that the "simplicity" created more problems than it solved – a touch interface was preferred by users.

Design Implications
Users often wanted to see other medication options after being recommended one. The “Other Options” page had no option to select the recommended dose or return to the previous screen.
  • Error recovery needed to be improved; allow for
    exploration of the UI without penalty
  • Streamline the user flows and make each path clear
  • Eliminate the “other options” page
At times, users skipped (and did not read) the text beneath the headline that told them what each button did. Users had no issue understanding and rating their pain from the scenario on a 0–3 point scale.
  • May need to intentionally shift UI elements to prevent mindless tapping
  • Pain scale is a good choice
  • Explicitly telling a person which button to press may not be necessary
When instructed to use the override button, several users often could not find it and tried to press the “yes” button in response to the onscreen prompt “Do you wish to override?”
  • Only show “override” when necessary
  • Improve visibility of active vs. inactive buttons
  • Revise flow again to find out where a prompt can be responded to in alternate ways
Older participants were generally quicker and did not have to backtrack as frequently.
  • Clarity is good and can be used properly by people of all backgrounds, ensure that this is continued throughout the rest of the UI
  • Need to improve error recovery flows
Form and Hardware
While our user testing was focused on the user experience of the UI, we were also able to gather some initial impressions around the physical design of the hardware. Since the product was intended to be set up in a prominent area of a patient's home so they would be able to see the alerts on the UI, ensuring that the physical design was visually appealing enough that it was not simply hidden away in some corner (defeating the purpose of alerts) was integral. Based on our usability testing and stakeholder feedback, some physical improvements we made included:
Issue / Concern
Initial placement of the screen facing straight up created issues of readability if a user was bedridden or not tall enough.
Reworked the physical panel where the screen is located to have a 45° angle to the base to allow for use from all directions. Screen at a more aggressive angle (e.g. 90°) was not ideal due to lateral movement of the device and needing to stoop down if device is located in a low position.
Pill dispensing area (the 'mouth') was too flat, requiring a user to hunt around for the pill in the dispensing area.
Pill dispensing area made more sloped and concave with a reduced 'lip' to enhance the 'scooping' motion of picking up the pill and allow for easier visual identification of the dispensed pill.
User Interface and User Flow
Working through the UI improvements that came out of our usability testing and multiple revisions of the UI flow with our stakeholders, some changes/improvements we made included:
  • Moved away from numeric pain scores to more patient experience-oriented pain scores backed by clinical research – we used the 'Wong-Baker Pain Scale' that is more explanatory and mitigates some of the phenomena of patients 'under-rating' their actual pain.
  • Moved away from a non-touch screen UI navigated by physical buttons to a full-touch UI as the complexity of the pain scales and overall flow (including "override' functionality) were not achieving high enough success with physical buttons.
  • Override functionality was revised to only be shown when necessary to increase its importance when such functionality is needed and/or available.
  • Error recovery was fleshed out and the flows were reworked to be always recoverable until the user confirms they want to dispense a pill.
  • Eliminated "Other Options" page and removed/integrated said features into the flow where appropriate instead.
Final Prototype
Form and Hardware
a video preview of a banking app
Internal Dispenser Design

The internal prototype dispensing mechanism was also fully built and prototyped both from hardware and software perspectives. This design is static, but could be adapted in the future to have a removable or disposable "tray" of pills for quick set-up at the pharmacy.

Fully-functional physical prototype

This is a fully-functional UI (per UI flow below) built on an Arduino Mega 2560 and off-the-shelf hardware in native C++. The dispensing mechanism is also functional for one full loaded rotation.

User Interface, User Flow & Content Design
a video preview of a banking app
High-level overview of "golden path" experience.

The UI is split into two primary paths – a "time to dispense" path and a "not time to dispense" path. We optimised for clarity and error recovery at every stage.

a video preview of a banking app
Detail - "time to dispense" flow

We use a Wong-Baker pain score to capture the most accurate yet personalised experience of a patient's pain, avoiding the pitfalls of "relative" experience of pain (e.g. a basic numerical score, where patients regularly under-report their pain). We keep it clear and prioritise patient agency. We're not able to make recommendations, but center the experience around reflexivity.

a video preview of a banking app
Detail - "not time to dispense" flow

We have both "soft" lockouts and "hard/full" lockouts. Hard lockouts are exclusively reserved for scenarios where it is physically unsafe to take another opioid. "Soft" lockouts can always be overridden in the case where a patient is in pain and does require more medication.

Reflection & Next Steps
In reflection, more user testing, especially with the finalised user interface, could have been conducted to ensure that the revised interface was less conducive to mistakes, could facilitate increased error recovery, increase clarity, and validate the idea that the touch screen was a better choice over physical buttons. With this testing, parts of the user flow could be further improved or validated, including aspects of content designs, decision trees, and button placement.

In terms of expanding the current functionality of the device, there are also a number of things to be addressed. Feasibility, device ownership dynamics, and mass-market loading techniques, among many others, must be fleshed out before this product can reach the market. For example, loading techniques such as pre-filled "wheels" that could be loaded like a cartridge by a pharmacist must be explored if this is to be used in any sort of mass-market setting. Ownership dynamics must also be explored: is this a device that a patient would rent from a hospital, pharmacy, or buy outright? Would this be paid for by insurance, and what motives are there for insurers to pay for such a device? Other not-so-obvious things such as how the device would be cleaned between different patients must also be explored, along with managing critical reception from medical professionals and administrative bodies such as the FDA– especially since this device is built to work with controlled substances.

As this is a proof-of-concept and not yet a production-ready, FDA-approved device, some of these limitations are understandable. However, this project is valuable more broadly in exploring and attempting to address these issues around outpatient opioid use (and the root causes of abuse), education, and dispensing practices.
Next Steps
Since this project was concluded, this project has registered as a University of Washington CoMotion Invention (#48371) and a company (Recap Medical, est. 2019) has been founded by our advisor for further development of the product in this problem space.