Browse docs
Docs · dashboard

Compulsive app checking

A quiet signal on the patient page for whether someone is opening Pippa compulsively. It counts check-only opens rather than minutes, and it is a prompt to ask a question in session, never a finding.

On this page

Repeat weighing and mirror-checking have an app analogue: opening the thing again and again, looking, and closing it without doing anything. This signal tries to notice that.

It sits on the patient page as a low-key tier — none, low, or medium — and it never escalates, never alerts, and never appears on the roster as a risk flag.

App checkinglow
Check-only opens are up on this patient's own baseline.
14 of the last 14 days had at least one open that ended without a log. Four of those were after 11pm.
low4 night daysbaseline 3/day → 9/day
How the signal appears on the patient page. A tier and a sentence, sitting where you are already looking. It never reaches the roster and never sends anything.

Why it does not count minutes

Time-in-app is a bad measure here specifically, for two reasons.

The first is that Pippa is built to hold attention. There is a pet, a room to decorate, music, and a coin economy. Minutes in app largely measure how well our own design works, which is not information about your patient.

The second is sharper: a forty-minute journaling session and forty one-minute panic checks look identical in minutes, and they are clinically opposite.

So the signal counts check-only opens — foregrounds that ended without the patient writing anything. The behaviour of interest is the looking, and the tell is that it produced nothing and had to be repeated.

This mirrors how the product already treats exercise: compulsivity predicts outcome, amount does not.

What it looks at

Recent check-only opens against that patient's own baseline, plus how many days ran over it, plus night-time opens. Someone who was already checking compulsively when they enrolled would otherwise look fine forever against their own high baseline, so there is an absolute floor too.

What it honestly cannot do

This matters more than the signal itself, so it is worth stating plainly. The limits are written into the code rather than discovered later:

  • A foreground is not an intention. Notification taps, returning from the camera, and a phone call ending all produce opens.
  • It cannot tell anxious checking from a delighted visit to the pet. The cozy room is designed to be revisited.
  • One device only. A second phone is invisible.
  • It cannot see other apps. The patient may be compulsively checking a different tracker entirely, and this will show nothing.
Never state it as a finding

This is a prompt to ask a question in session. "I noticed you've been opening the app a lot lately, how's that been?" is the use. "The app says you're checking compulsively" is not — you would be telling a patient something about themselves that the data cannot actually support, and in this population that lands hard.

Why it never alerts

An escalation would require confidence the signal does not have. Given how many innocent things produce an open, a tier that pushed a notification at you would be wrong often enough to be ignored within a fortnight, and an ignored alert is worse than none.

It is there when you are already looking at the patient, which is when it is useful.

For a collaborative plan to reduce reliance while preserving care, see Stepping down Pippa support.

Was this page helpful?