RemindCare AI

UX Design & Prototyping

1-Day Hackathon

June 2025

I designed and prototyped an adaptive reminder system that tailors messaging to each patient's no-show risk. Built in one day with two product managers who brought real clinic data to the table. After presenting at the hackathon, the notifications team began exploring how to pull the adaptive messaging approach into their product.

The Problem

Missed medical appointments cost healthcare systems $150–200 per missed visit and disrupt continuity of care. But the real problem isn't forgetfulness. Standard reminders treat every patient the same, even though people miss appointments for very different reasons: transportation access, health literacy, financial stress, or simply the chaos of daily life. I wanted to explore whether adaptive notifications could change that.

Understanding the Problem Space

To design something that could actually work, I needed to understand why patients miss appointments in the first place. I dug into predictive models ^[1], factors behind high no-show rates across patient populations ^[2], and how message framing impacts behavior ^[4].

The research showed that no-shows are predictable based on factors like appointment history, insurance type, and access barriers. Models using multiple data sources can forecast risk with meaningful accuracy ^[3]. More importantly, how you communicate matters as much as when you communicate ^[4].

How it works

RemindCare AI combines multiple data signals to build a risk profile for each appointment, then tailors the reminder strategy accordingly.

Assess Risk

Analyzes patient history (past no-shows, late cancellations), demographics (age, insurance type), and environmental factors (weather, traffic, commute) to calculate a no-show probability score. Also incorporates clinic-level metrics like visit type and location data.

Adapt Messages

Based on the risk level, the tone, timing, and content of reminders change. Low-risk patients get simple confirmations. High-risk patients receive empathetic messages with proactive help like transportation assistance.

Intervene Early

Reminders are sent at 7 days, 3 days, and 1 day before the appointment. Each touchpoint escalates if the patient hasn't confirmed, giving multiple chances to reschedule before they no-show.

The Experience

The same appointment gets a completely different reminder experience depending on the patient's risk profile. Use the controls below to see how risk factors change the messaging strategy.

Patient Scenario

Patient Presets

Visit Type

Location

Patient History

No-Shows (12 mo)

Late Cancels (12 mo)

Demographics

Age

Insurance

Environment

Weather

Traffic

Commute

Risk Assessment

No-show risk

Low Risk

Contributing factors

Visit type (Example data)+7%

Message Preview

RemindCare

Wed, May 14

Hi Sarah, your visit with Dr. Smith is 📅 Tuesday, May 21 at 2:00 PM. We're looking forward to seeing you. Reply: 1 to confirm 2 to reschedule 3 to cancel

Sun, May 18

Hi Sarah, just a reminder, your visit with Dr. Smith is in 3 days (📅 Tuesday, May 21 at 2:00 PM). Reply: 1 to confirm 2 to reschedule 3 to cancel

Tue, May 20

Hi Sarah, your visit with Dr. Smith is tomorrow at 2:00 PM. ⛅ Forecast: clear weather, traffic looks light (~25 min). Reply: 1 to confirm 2 to reschedule 3 to cancel

Type a message

Process & Reflections

How I worked

I spent the morning on research, then used AI to build both the risk model and this interactive prototype. My PM teammates provided real clinic data, visit-type rates, cancellation patterns, insurance breakdowns, which kept everything grounded. I'm not a developer, so AI let me focus on UX decisions rather than code.

Why AI changed the approach

Instead of creating dozens of static mockups for every scenario, I built a tool where I defined the rules and let others explore the possibilities. That meant going from idea to interactive demo in a single day, and showing what something feels like, not just what it looks like.

What I'd do next

Test the messages with real patients. The copy is research-backed, but I don't know how someone with three prior no-shows actually feels reading "we're counting on seeing you." I'd also explore channel preferences beyond SMS and enrich the risk model with social determinants of health.

Potential Impact

No production data yet, but the research is promising: personalized message framing alone can reduce no-show rates by 4–8% ^[4], and risk models can correctly flag high-risk appointments over 80% of the time ^[3]. Even a modest reduction means recovered revenue, better patient outcomes, and more efficient provider time.

[1] Kheirkhah, P., Feng, Q., Travis, L.M., Tavakoli-Tabasi, S., & Sharafkhaneh, A. (2016). Prevalence, predictors and economic consequences of no-shows. BMC Health Services Research, 16, 13.

[2] Samuels, R.C., Ward, V.L., Melvin, P., Macht-Greenberg, M., Wenren, L.M., Yi, J., Massey, G., & Cox, J.E. (2015). Missed Appointments: Factors Contributing to High No-Show Rates in an Urban Pediatrics Primary Care Clinic. Clinical Pediatrics, 54(10), 976–982.

[3] Huang, Y., & Hanauer, D.A. (2014). Patient No-Show Predictive Model Development using Multiple Data Sources for an Effective Overbooking Approach. Applied Clinical Informatics, 5(3), 836–860.

[4] Berliner Senderey, A., Kornitzer, T., Lawrence, G., Zysman, H., Hallak, Y., Ariely, D., & Balicer, R. It's how you say it: Systematic A/B testing of digital messaging cut hospital no-show rates.