Accessible Medication Reminder

Designed first for the actual user

Most medication apps are designed for a 30-year-old product manager and then “made accessible” as a retrofit. This concept inverts the order: the primary user is an 80-year-old with reading glasses, a tremor, and a phone their daughter set up. Type is enormous by default, contrast is high, every screen carries exactly one action, targets are forgiving of imprecise taps, and the entire flow works by voice — “medication taken” spoken aloud confirms a dose, and confirmations are echoed back in speech and haptics. If it works for this user, it works for everyone; the reverse has never been true.

The one notification that must not be missed

A missed social notification costs nothing; a missed dose can cost a hospital stay. So the reliability engineering that most apps skip is this product’s core. Mobile notifications are best-effort by design — silenced modes, battery optimizers that kill background apps, timezone and clock-change edge cases all eat reminders silently. The concept treats delivery as a layered system: alarm-class notifications where the OS allows them, a full-screen takeover with sound for critical doses, an unmissable repeat schedule until confirmed, and — with consent — escalation to a caregiver after a defined number of unconfirmed reminders. Every reminder’s delivery and confirmation is logged, and the app tests its own delivery path, warning plainly when a device setting threatens it: “Your phone may silence reminders overnight. Fix this together with the person who helps with your phone.”

Caregiver visibility, on the user’s terms

Nothing is shared by default. The user (or the user with help) explicitly invites a caregiver, and the app shows in plain language exactly what that person can see — confirmations and misses, never more. The invitation is revocable in one obvious place. Dignity is a feature: the user remains the account’s owner, not its subject.

Deliberately no AI

A reminder must be boring and exact. There is no model between the schedule and the alarm, no generated encouragement, no adaptive timing that a user can’t predict, and no chatbot answering medication questions — those belong to a pharmacist. Predictability is the product; AI would only add variance where variance is harm.

Evaluation plan

• Usability sessions with older adults and low-vision users, including screen-reader and voice-only passes — task completion without sighted or younger help as the bar
• A delivery-reliability harness across common devices and battery-saver settings, with a zero-silent-loss target for critical-dose reminders
• Comprehension check on the consent flow: can users accurately say what their caregiver sees (hypothesis: all of them — anything less is a redesign)