45
Case study
The Connect Records Study
The hypothesis was simple: make the codeless access path more visible and fewer patients would call support. I ran a between-subjects study across 45 participants to test it. The data said the hypothesis was wrong. That turned out to be more useful than a win would have been.
tl;dr
- All three variants performed well on clarity and navigability.
- The current Control design outperformed both redesigns on post-task confidence and expectation alignment while keeping call likelihood low.
- The Preliminary decision-first variant performed worst across nearly every measure.
- The study disproved its own central hypothesis: patients were not simply missing the codeless pathway.
- The findings redirected the team back toward the telephony data and the access issues underneath it.
What We Learned
What the Study Was Designed to Test
The Connect Your Records flow asks patients to link their health records to their AdventHealth account using a 15-character code found on their After Visit Summary. Patients who do not have the code can proceed without it, but that pathway was not always obvious, and the hypothesis was that patients were calling support simply because they could not find it.
Three variants were tested: the current Control, a Side-by-Side version that gave both pathways equal visual weight, and a Preliminary decision-first version that asked patients up front whether they had a code before showing them the form. Each participant saw one variant and completed post-task confidence, alignment, and call-likelihood measures on a 1–7 scale.
Variant comparison
Where participant attention concentrated
Control
AdventHealth
Connect Your Records
Enter your access code and date of birth.
Side-by-Side
AdventHealth
Enter your access code
Both pathways are visible with equal weight.
Preliminary
AdventHealth
Do you have a personal activation code?
Patients are asked to decide before they have context.
Confidence
The Control held its ground
The Control Held Its Ground
The current production design, which the team expected to be the weakest performer, earned the highest post-task confidence scores and the strongest expectation alignment of all three variants. Participants described it as intuitive and easy to follow.
They did not struggle to find the codeless pathway. They found it and used it without significant hesitation. That directly challenged the underlying assumption that visibility alone was the source of support demand.
The Side-by-Side Was a Modest Improvement in One Dimension
Presenting both pathways with equal visual weight produced the lowest overall call likelihood of any variant. That was meaningful. But it trailed the Control on post-confidence and expectation alignment, suggesting that surfacing both options simultaneously introduced some ambiguity about which path applied to the participant’s situation.
A cleaner visual treatment could move one metric in the right direction without becoming the most trusted version of the experience overall.
Support likelihood
The Preliminary variant created the problem it was trying to solve
The decision-first design, asking patients to declare up front whether they had a code before showing them anything else, produced the worst outcomes across the board: highest call likelihood, lowest confidence, lowest expectation alignment, and the longest time-on-study. Participant responses pointed to why: being asked to make a decision before understanding the context introduced uncertainty rather than reducing it.
Expectation alignment
The real finding was upstream of the interface
The Real Finding: This Was a Mental Model Problem, Not a Visibility Problem
The most important output of this study was what it ruled out. Patients across all three variants navigated the flow successfully. They were not missing the codeless pathway. What they lacked was a clear understanding of what the code was, where it came from, and what connecting records actually meant for them.
That gap, between system terminology and patient mental model, is not something a button label or layout change can close. It pointed back to the telephony findings: the root cause was upstream of the interface.
Outcome of the Research
The study produced three near-term recommendations: dive deeper into the Five9 call data to pinpoint the specific call drivers tied to the Connect Records flow; reframe the codeless pathway so that proceeding without a code feels normal and expected rather than like a workaround; and add Amplitude analytics tags to the flow to capture actual patient behavior in production rather than inferring it from prototype studies.
Recommendations
- Deep dive into Five9 call drivers tied specifically to Connect Records.
- Make “no code” feel normal and safe rather than exceptional.
- Add Amplitude tags so production behavior can be observed directly.
Rather than moving forward with a variant redesign, the team redirected attention toward the upstream access-dependency issues. The study was built to test whether visibility could reduce call volume. The data said no, and that clarity was the actual outcome worth shipping.
Continue Exploring
Contact
Want to talk through the Connect Records study?
The variant design, how the hypothesis shifted mid-study, or what it looks like when a study disproves itself: happy to walk through any of it.